Design and Construction of the DEAP-3600 DarkMatter Detector

P.-A. Amaudruzl, M. Baldwinj, M. Batygovd, B. Beltrana, C. E. Binaa,D. Bishopl, J. Bonattf, G. Boormani, M. G. Boulayf,c, B. Broermanf,

T. Bromwichh, J. F. Buenoa, P. M. Burghardtk, A. Butcheri, B. Caif, S. Chanl,M. Chenf, R. Chouinarda, S. Churchwellh, B. T. Clevelandg,d, D. Cranshawf,

K. Deringf, J. DiGioseffof, S. Dittmeierl, F. A. Duncan†,g,d, M. Dunfordc,A. Erlandsonb,c, N. Fatemighomii, S. Florianf, A. Flowerf, R. J. Fordg,d,

R. Gagnonf, P. Giampaf, V. V. Golovkob, P. Gorela,g,d, R. Gorneac, E. Gracei,K. Grahamc, D. R. Granta, E. Gulyevl, A. Halli, A. L. Hallina, M. Hamstraf,c,

P. J. Harveyf, C. Hearnsf, C. J. Jillingsg,d, O. Kamaevb, A. Kempi,M. Kuzniakf,c, S. Langrockd, F. La Ziai, B. Lehnertc, O. Lig, J. J. Lidgardf,

P. Liimataineng, C. Liml, T. Lindnerl, Y. Linnl, S. Liua, P. Majewskij,R. Mathewf, A. B. McDonaldf, T. McElroya, K. McFarlaneg, T. McGinn†,f,

J. B. McLaughlinf, S. Meadl, R. Mehdiyevc, C. Mielnichuka, J. Monroei,A. Muirl, P. Nadeauf, C. Nantaisf, C. Nga, A. J. Noblef, E. O’Dwyerf,C. Ohlmannl, K. Olchanskil, K. S. Olsena, C. Ouelletc, P. Pasuthipf,

S. J. M. Peetersh, T. R. Pollmannd,f,k, E. T. Randb, W. Rauf, C. Rethmeierc,F. Retierel, N. Seeburni, B. Shawl, K. Singhraol,a, P. Skensvedf, B. Smithl,

N. J. T. Smithg,d, T. Sonleyf, J. Soukupa, R. Stainforthc, C. Stonef,V. Stricklandl,c, B. Surb, J. Tanga, J. Taylori, L. Velocef,

E. Vazquez-Jaureguig,d,e, J. Waldingi, M. Wardf, S. Westerdalec, R. Whiteh,E. Woolseya, J. Zielinskil

aDepartment of Physics, University of Alberta,Edmonton, Alberta, T6G 2R3, Canada

bCanadian Nuclear Laboratories Ltd., Chalk River Laboratories,Chalk River, K0J 1P0 Canada

cDepartment of Physics, Carleton University,Ottawa, Ontario, K1S 5B6, Canada

dDepartment of Physics and Astronomy, Laurentian University,Sudbury, Ontario, P3E 2C6, Canada

eInstituto de Fısica Universidad Nacional Autonoma de Mexico,Apartado Postal 20-364, Mexico D. F. 01000

fDepartment of Physics, Engineering Physics, and Astronomy, Queen’s University,Kingston, Ontario, K7L 3N6, Canada

gSNOLAB, Lively, Ontario, P3Y 1M3, CanadahDepartment of Physics and Astronomy, University of Sussex,

Sussex House, Brighton, East Sussex BN1 9RH, United KingdomiDepartment of Physics, Royal Holloway, University of London,

Egham Hill, Egham, Surrey TW20 0EX, United KingdomjRutherford Appleton Laboratories, Swindon SN2 1SZ, United Kingdom

kDepartment of Physics, Technische Universitat Munchen,80333 Munich, Germany

lTRIUMF, Vancouver, British Columbia, V6T 2A3, Canada

†Deceased.

Preprint submitted to Astroparticle Physics April 11, 2018

arX

iv:1

712.

0198

2v2

[as

tro-

ph.I

M]

10

Apr

201

8

Abstract

The Dark matter Experiment using Argon Pulse-shape discrimination (DEAP)

has been designed for a direct detection search for particle dark matter using

a single-phase liquid argon target. The projected cross section sensitivity for

DEAP-3600 to the spin-independent scattering of Weakly Interacting Massive

Particles (WIMPs) on nucleons is 10−46 cm2 for a 100 GeV/c2 WIMP mass

with a fiducial exposure of 3 tonne-years. This paper describes the physical

properties and construction of the DEAP-3600 detector.

Keywords: dark matter, WIMP, liquid argon, DEAP, SNOLAB, low back-

ground

2

Contents

1 Introduction 4

1.1 Detector Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.2 Design Realization . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Material Selection 10

2.1 Detector Simulation . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.2 Material Assay Techniques . . . . . . . . . . . . . . . . . . . . . . 14

2.3 Material Production and Quality Assurance . . . . . . . . . . . . 19

3 Cryogenic System 23

3.1 Purification System . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.2 Neck Seal Incident . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4 Inner Detector Construction 30

5 Light Detection Systems 39

5.1 Photomultiplier Tubes . . . . . . . . . . . . . . . . . . . . . . . . 39

5.2 Neck Veto System . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.3 Calibration Systems . . . . . . . . . . . . . . . . . . . . . . . . . 45

6 Electronics 50

6.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

6.2 Software and Data Rate Reduction . . . . . . . . . . . . . . . . . 55

6.3 Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.4 Database and Data Flow . . . . . . . . . . . . . . . . . . . . . . . 56

7 Detector Infrastructure 57

8 Safety 60

8.1 Over-pressure Protection . . . . . . . . . . . . . . . . . . . . . . . 60

8.2 Oxygen Deficiency . . . . . . . . . . . . . . . . . . . . . . . . . . 60

9 Summary 61

3

1. Introduction

The origin of dark matter in the universe is one of the most important

questions in particle astrophysics. A well-motivated dark matter candidate is

the Weakly Interacting Massive Particle (WIMP), which is predicted naturally

in supersymmetric extensions of the Standard Model [1, 2]. To date, WIMPs

remain undetected in laboratory-based searches [3, 4, 5]. Direct detection ex-

periments aim to measure energy depositions of order 100 keV and below, and

suppress backgrounds to the level of less than one event per tonne per year to

create a signal search region free from backgrounds.

DEAP-3600 (Dark matter Experiment using Argon Pulse-shape discrimina-

tion) has been designed to perform a direct WIMP dark matter search using

3600 kg of liquid argon (LAr) as a target. DEAP-3600 is located 2 km un-

derground (6000 meters water equivalent overburden) at SNOLAB in Sudbury,

Ontario, Canada [6] and builds on the technology developed on the DEAP-1

prototype containing 7 kg of LAr [7]. The projected sensitivity to the spin-

independent WIMP-nucleon cross-section is 10−46 cm2 for a WIMP mass of

100 GeV/c2 [7, 8]. Careful design and construction of the DEAP-3600 detector

have reduced the predicted backgrounds in the search region of interest to less

than 1 event in a fiducial exposure of 3 tonne-years.

Central features of the detector design include:

1. Single phase target: a monolithic inner volume of LAr allows minimal de-

tector material to be in contact with the argon target, which can be made

very radiopure. Particle energy deposition in the single phase LAr pro-

duces scintillation photons. The LAr scintillation time structure provides

discrimination between WIMP-induced nuclear recoil events and electro-

magnetic background events [7]. In addition to event characterization,

event locations can be reconstructed using the detected scintillation sig-

nals.

2. Use of an acrylic cryostat: acrylic can be produced in a very controlled

and radiopure fashion. As a hydrogenous material, it is a good neutron

4

shield and possesses useful optical, mechanical, and thermal properties.

The capacity for a large thermal gradient across acrylic allows for light

guides of reasonable length with a cryogenic inside surface and an outside

surface coupled to near-room-temperature photomultiplier tubes (PMTs)

for signal readout.

3. Radiopure raw materials: the detector is constructed of ultra-clean ma-

terials. A quality assurance and testing program during procurement,

manufacture, and construction provides careful control and inventory of

radioactive contaminants.

4. Electronics optimized for LAr scintillation detection: excellent single pho-

toelectron response and digitization over 16 µs enable full exploitation of

the scintillation time structure for particle identification, with low dead-

time and manageable data rates.

1.1. Detector Overview

A schematic view of the DEAP-3600 detector is shown in Figure 1, and the

main design parameters are summarized in Table 1. The inner detector includes

the acrylic cryostat, neutron shielding materials, and the array of PMTs that

view the LAr volume. The material selection and assay campaign for all detector

materials are described in Section 2. The LAr purification system is described

in Section 3 and inner detector components are described in Section 4.

The cryostat consists of a 5-cm-thick spherical acrylic vessel (AV), 85 cm

in inner radius which can contain 3600 kg of LAr. Acrylic light guides (LGs),

45 cm in length, are directly bonded to the vessel. The LGs couple the AV to the

PMTs and provide neutron shielding. Interspersed between the LGs are filler

blocks comprised of layers of high-density polyethylene and polystyrene, which

complete the neutron-shielding sphere. The inner surface of the AV is coated

with a 3-µm-thick layer of the organic wavelength-shifter 1,1,4,4-tetraphenyl-

1,3-butadiene (TPB, C28H22), deposited in situ [9], to convert the argon scin-

tillation light into the visible wavelength region for transmission through the

LGs to the PMT array. The target volume is viewed by 255 8-inch-diameter

5

Table 1: DEAP-3600 detector design parameters for a 3-tonne-year exposure.

Parameter Design Specification

Sensitivity at 100 GeV/c2 10−46 cm2

Backgrounds target < 0.6 eventsNominal region of interest (ROI) 120–240 photoelectronsNominal analysis threshold 15 keVee

Fiducial mass, radius 1000 kg, 55 cm

Number of HQE inner detector PMTs 255Light yield 8 photoelectrons/keVee

Nominal position resolution at threshold 10 cm1

Total argon mass, radius 3600 kg, 85 cmWater shielding tank diameter x height 7.8 m x 7.8 mNumber of Cherenkov veto PMTs 48

Hamamatsu R5912-HQE high quantum efficiency (HQE) PMTs manufactured

with low radioactivity glass. The PMTs and associated readout electronics are

described in Sections 5 and 6, respectively.

The inner detector is housed in a stainless steel pressure vessel, comprising

a spherical shell and the outer neck. Access to the inner detector volume is

through a steel and acrylic neck that couples to the AV. This neck contains

a cooling coil which uses liquid nitrogen (LN2) to cool the LAr. A glove box

interface at the top of the neck allows insertion or extraction of equipment in a

radon-free environment. The outer steel shell is immersed in a 7.8-m-diameter

shield tank filled with ultra-pure water and instrumented with 48 Hamamatsu

R1408 PMTs serving as a muon veto. The pressure vessel and shield tank are

described in Section 7. Safety systems for maintaining a large, liquid argon

volume in an underground environment are described in Section 8.

A charged particle traversing the LAr target in DEAP-3600 loses energy

through ionization. Scintillation emission is produced through the excitation or

ionization of neutral argon atoms and the resulting creation of unstable argon

dimers. These dimers decay with a characteristic singlet or triplet state lifetime

of 6 ns or 1.5 µs, respectively, and the ratio of production into these two states

1Updated simulations predict an improved position resolution.

6

Figure 1: The DEAP-3600 detector design showing the acrylic vessel, lightguides, filler blocks, steel shell, neck, and glove box. Not shown are the wave-length shifting coating over the interior of the acrylic vessel and the surroundingmuon veto water tank.

7

is dependent on the linear energy transfer of the incident radiation [10]. The

scintillation efficiency for nuclear recoil-induced events is quenched by approxi-

mately 0.25 with respect to events induced by electron recoils [11]. The resulting

photon spectrum from argon dimer decay is in the vacuum ultraviolet (VUV)

region, peaked at 128 nm [12]. This is lower in energy than the first atomic ex-

cited state of neutral argon, permitting the photons to travel through the argon

without absorption. When these VUV photons reach the inner surface of the

AV, they are absorbed by the TPB coating and re-emitted in the visible region,

peaked at approximately 420 nm [13], near the peak quantum efficiency of the

PMTs [14]. The wavelength-shifted photons are transmitted through the acrylic

light guides to the PMTs. The output signal is split into high- and low-gain

channels, amplified, and shaped with custom electronics.

1.2. Design Realization

The design required research and development in many areas of detector

composition and construction. As this is the first use of a large acrylic cryostat,

the mechanical properties of acrylic and acrylic bonds at liquid argon temper-

ature (87 K) were measured at Los Alamos National Laboratory [15]. Large

thermal stresses from differential contraction during operation and cooling were

modeled with finite element analysis (FEA) and tested in detail. A technique

for bonding the densely-packed LGs onto the acrylic sphere that is cryogenically

and mechanically robust, clean, geometrically precise, and optically transpar-

ent was developed. A sanding resurfacer robot was built of radiopure materials

with the ability to remove contamination from the entire inner acrylic surface.

A method of large-scale, thin-film deposition, used to coat the inner acrylic

surface with the wavelength shifter was developed and deployed [9]. Lastly, an

argon purification and radon-removal system was developed.

The aim throughout design and construction was to minimize internal and

external background interactions that could mimic a dark matter signal. These

include: (i) backgrounds from electromagnetic events, particularly the beta de-

cay of the naturally-occurring 39Ar; (ii) radon and radon progeny in the argon

8

volume; (iii) radioactivity at or near the inner acrylic surface; (iv) neutrons;

and (v) backgrounds associated with cosmic rays.

Mitigation of the large 39Ar background (approximately 1 Bq per kg of nat-

ural argon [16]) is accomplished using pulse-shape discrimination (PSD) of the

scintillation signal. This technique is very powerful in LAr and has a projected

discrimination power of 10−10 [7, 17] for DEAP-3600. The HQE PMTs are oper-

ated near room temperature to allow optimal performance and light collection,

upon which the PSD strongly depends.

The radon background mitigation strategy employs material Rn emanation

assay and selection, control of exposure to lab air during production, and surface

treatment after construction, in addition to the cryogenic purification of the

argon. Radon (222Rn and 220Rn) and progeny in the LAr target target itself

are reduced by carefully controlling the materials and construction of the argon

purification systems and inner detector.

The acrylic for the cryostat was chosen carefully, and its production was

monitored directly by the DEAP-3600 collaboration, minimizing exposure of

the acrylic precursors to radon and other contaminants. Construction of the

acrylic vessel was also highly controlled; however, the final assembly steps were

performed underground at SNOLAB, where the 222Rn level is approximately

130 Bq/m3. The robotic resurfacer was built to remove radio-contaminants

built up on the inner acrylic surface from radon surface deposition and diffusion

following the AV construction underground. To minimize the contribution to

surface backgrounds from the wavelength shifting coating applied after resur-

facing, the production of the TPB was monitored by the collaboration. Finally,

surface backgrounds are reduced in analysis by using position reconstruction

algorithms to reject events near the inner AV surface.

Neutrons produced internally in the detector are controlled with radiopure

materials to reduce their production rate and with hydrogenous shielding to

thermalize those that are produced near the LAr. The primary source of neu-

trons, the borosilicate PMT glass, is effectively moderated in the acrylic LGs

and polyethylene filler blocks. Neutrons from the rock wall at SNOLAB are

9

moderated by the water-filled shield tank. Cosmic ray muons, with a flux of

0.27 m−2 day−1 [6], are tagged with the Cherenkov light detection veto system

to reject the cosmogenic neutrons they produce.

Table 2 summarizes the background goals for a 3 tonne-year fiducial exposure

in the energy ROI of 120–240 photoelectrons, corresponding to a nominal energy

ROI of 15 to 30 keVee.

Table 2: Targeted number of events in the energy ROI, 120 < photoelectrons <240, with a 3-tonne-year fiducial exposure. Fiducialisation assumes both a 10 cmposition resolution for surface events and 50% nuclear recoil acceptance frompulse shape discrimination.

Source Events in Energy ROIFiducial Eventsin Energy ROI

Neutrons 30 < 0.2Alphas (surface) 150 < 0.2Betas/gammas (39Ar dominated) 1.6 × 109 < 0.2

Sum < 0.6

2. Material Selection

Detector materials were carefully selected for radiopurity and optical proper-

ties to maximize the dark matter detection sensitivity. An extensive radiopurity

assay campaign was performed in combination with Monte Carlo simulations to

ensure adherence to the targeted background budget.

2.1. Detector Simulation

The DEAP-3600 detector simulation uses RAT [18], a software framework for

simulation and analysis of liquid scintillator experiments, which uses Geant4 [19]

version 4.9.6 and ROOT [20] version 5.34 libraries. Customized versions of RAT

are currently used by the MiniCLEAN, SNO+, and DEAP-3600 collaborations.

Simulations have been used extensively for the definition of radiopurity re-

quirements for detector materials, background rejection studies including shield-

ing optimization and position reconstruction, light yield optimization, activity

10

requirements for calibration sources, and studies of electronics or trigger-related

biases and other systematic effects.

The simulation implements the full, as-built detector geometry, including

the SNOLAB Cube Hall cavern in which the DEAP-3600 detector is located.

A number of optical parameters of the simulation are defined by ex-situ mea-

surements, which include the wavelength-dependent light attenuation length of

acrylic [21], wavelength-dependent reflectance of the diffuse2 and specular re-

flectors surrounding the AV, as well as the alpha scintillation properties of TPB,

including light yield [23], the scintillation time profile information, and its tem-

perature dependence [24, 25]. Optical photons are fully propagated in RAT and

the process has been validated using DEAP-1 data [26].

The following extension packages are used in the simulation in addition to

the standard Geant4 physics processes:

• A detailed simulation model for nuclear recoils on arbitrarily rough sur-

faces was developed [27] to investigate backgrounds originating from the

inner surface of the AV. The model combines explicit surface roughness

implementation in the geometry with a Geant4 extension, available as one

of its extended examples (TestEm7), which contains all physics relevant for

multiple inter-atomic and alpha scattering in the 10 keV–10 MeV energy

range [28, 29]. It has been extensively benchmarked against SRIM [30]

with respect to nuclear straggling and implantation, as well as backscat-

tering.

• Argon scintillation is simulated based on the model from [31]. The NEST

(Noble Element Scintillation Technique) model [32], developed primarily

for xenon and used by LUX, EXO, and XENON100 collaborations, is

optionally available in RAT. SCENE measurements to calculate nuclear

recoil quenching factors and PSD distributions are also implemented in

RAT.

2Data kindly contributed by Martin Janecek (LBNL), as measured with the apparatusdescribed in [22].

11

• The hadronic physics models relevant for muon- and gamma-induced neu-

tron simulations recommended in [33, 34] have been adopted. At energies

below 20 MeV, high precision data-driven neutron models provided by

Geant4 are used.

• The energy spectrum and rate of neutrons from inner detector components

is calculated using the SOURCES-4C code [35] and cross-checked using

NeuCBOT [36].

Based on the background targets in Table 2, material selection efforts were

focused on mitigating backgrounds from alphas and neutrons. The activity from

39Ar beta decays dominates the overall rate unless special, sequestered argon

sources depleted in cosmogenically-produced 39Ar are employed [37, 4]. These

events, however, are readily removed from the analysis using PSD. Each poten-

tial contributing source to the alpha and neutron backgrounds is normalized to

a total of 0.2 background events in a 3-tonne-year fiducial exposure to set target

values.

In the background model, the leading alpha backgrounds originate from

contamination in the wavelength shifter and acrylic bulk. While traversing

these regions, alphas can lose energy, reducing the typically high energy signal

down into the WIMP search ROI. The targeted alpha background activity is

set by normalizing to 0.2 background events in a 3-tonne-year fiducial exposure

and assuming a 10 cm position reconstruction resolution at threshold, resulting

in a factor of 103 reduction in surface background rate in analysis. The targeted

activity of the 238U, 232Th, and 210Pb chains for TPB and acrylic is summarized

in Table 3 assuming secular equilibrium and 1 alpha from 210Pb per 238U decay.

The main sources of internal neutron backgrounds come from the production

of neutrons through (α, n) reactions and spontaneous fission in the detector

materials. The dominant components contributing to the neutron backgrounds

in the simulation model are summarized in Table 4, along with the targeted

radiopurity values reported by normalizing to a total 0.2 background events

from neutrons in a 3-tonne-year fiducial exposure.

12

Table 3: Targeted activity of the 238U, 232Th, and 210Pb decay chains in TPBand AV acrylic contributing to the surface alpha background assuming secularequilibrium apart from 210Pb. Surface alphas can lose energy in the acrylic andTPB layer reducing the energy deposited in the LAr volume into the WIMPsearch ROI. The sources of alphas are normalized to a total of 0.2 backgroundevents in a 3-tonne-year fiducial exposure to report the targeted values.

Targeted Activity [µBq/kg]Component 238U 232Th 210Pb

TPB 5.7 8.8 4.0AV acrylic bulk 2.9 3.9 20.0

Table 4: Specific activity targets of the 238U, 235U, and 232Th decay chainsin detector materials contributing to the neutron background assuming secularequilibrium. Target values are reported by normalizing to a total of 0.2 back-ground events in the WIMP energy ROI with a 3-tonne-year fiducial exposurefrom neutrons emitted by (α, n) and spontaneous fission reactions.

Targeted Activity [mBq/kg]Component 238U 235U 232Th

PMT glass 82.8 72.0 47.2PMT ceramic 3530 — 960AV acrylic 0.02 0.09 0.08LG acrylic 0.12 0.19 0.16Filler blocks (polyethylene) 0.36 0.53 0.54PMT mount PVC 124.0 72.0 49.2Neck Steel 19.2 96.0 19.2Neck PMT glass 24300 — 11600

The radon emanation rate for argon-wetted materials (purification system,

detector inner-neck components) is constrained by the allowable alpha back-

ground on the inner AV source. For the region outside the AV but within the

steel shell, the allowable radon load is determined by assuming that progeny

collect on the outer AV surface approximately 5 cm from the inner AV surface,

using the conservative assumption that all progeny will stick to the detector

surfaces. The targeted radon emanation rates for inner detector components

to maintain the neutron background target are shown in Table 5. In practice

during detector operation, the steel shell region is purged with radon-scrubbed

13

boil-off nitrogen gas; the actual radon load in this region is should be that of

the purge gas.

Table 5: Targeted radon emanation rates for major inner detector materials andcomponents. Target values are reported by normalizing to a total of 0.2 back-ground events in the WIMP energy ROI with a 3-tonne-year fiducial exposurefrom neutrons emitted by (α, n) reactions by radon plated-out on cold inner-detector surfaces.

Radon Emanation Source Target Emanation Rate

PMT cables 0.047 [mBq/m]AV acrylic 5.5 [mBq/m2]LG acrylic 1.2 [mBq/m2]Filler blocks 0.7 [mBq/m2]PMT mount (PVC) 1.8 [mBq/m2]FINEMET PMT magnetic shielding [38] 2.0 [mBq/m2]Stainless steel shell 2.7 [mBq/m2]PMTs 0.4 [mBq/PMT]

2.2. Material Assay Techniques

To reach the radiopurity goals summarized in Tables 3, 4, and 5, extensive

low-background gamma assay and radon emanation measurement programs to

select materials, in addition to material and handling quality assurance pro-

grams, were developed.

SNOLAB has a well-established gamma assay program; a 200 cm3 high-

purity germanium well detector (Princeton Gamma-Tech Instruments, Inc.) [39]

was purchased and installed at SNOLAB to meet the assay requirements for the

DEAP experiment. An inventory of gamma assay measurements3 for the 238U,

232Th, and 235U decay chains which contribute to the neutron background from

major detector components is listed in Table 6 and from tools used in manufac-

turing components in Table 7. Material assays measuring the 234Th and 234mPa

gamma lines in the 238U decay chain often show different activities from what is

observed in the 226Ra, 214Pb, and 214Bi gamma lines. This discrepancy indicates

3Radiopurity database: https://deap-radiopurity.physics.carleton.ca/database/

14

that secular equilibrium is often broken between 230Th and 226Ra. 238U and

its progeny up to and including 230Th is referred to as the “238Uupper chain”,

while 226Ra and its progeny as the “238Ulower chain”, assuming secular equilib-

rium within each sub-chain. Assay results contributing to the electromagnetic

backgrounds (40K, 60Co) are measured, but not reported in Table 6. Uncertain-

ties arise from counting statistics and detection efficiency. A description of the

components can be found in Sections 4 and 5.

A new radon emanation measurement system constructed at Queen’s Uni-

versity, similar to that described in [41], was used to qualify and select detector

materials. Materials are loaded into a vacuum chamber, the chamber is evacu-

ated, and the material is then allowed to emanate into vacuum before the Rn

atoms are collected using a cold trap. The alpha decay rate from the Rn atoms

is measured to determine the initial radon emanation rate from the material.

Uncertainties in the emanation rate are due to system backgrounds, and the

efficiencies of trapping and detection. Typical backgrounds in the emanation

system are on the order of a single radon atom. Upper limits are set when there

is no signal above background. Table 8 lists results from the 222Rn emanation

of the main detector components and tooling used during fabrication. Limits on

the AV and LG acrylic emanation are both set at < 0.3 mBq/m2; the AV radon

emanation is based on the assayed uranium content and not a direct radon em-

anation measurement. The steel shell volume is purged using a boil-off nitrogen

gas system with a radon emanation rate of 5 µBq/kg [42].

Radon emanated from the resurfacer sanding robot, which contributes to

the surface alpha background through possible collection of the long-lived radon

daughter 210Pb while the sanding robot is deployed in the AV, was calculated

based on screening measurements of the individual components to be less than

20 mBq.

The 210Pb targets set by the maximum allowable contribution to (α, n) back-

grounds are below the sensitivity of most current assay techniques. A program

4Candidate material, not used in final construction.

15

Tab

le6:

Gam

ma

ass

ay

resu

lts

for

ma

jor

det

ecto

rco

mp

on

ents

.A

des

crip

tion

of

the

com

pon

ents

can

be

fou

nd

inS

ecti

on

s4

an

d5.

Act

ivit

ies

are

rep

ort

edw

ith

1-s

igm

au

nce

rtain

ties

.A

90%

con

fid

ence

lim

itis

pla

ced

wh

enth

em

easu

rem

ent

isb

elow

the

back

gro

un

dse

nsi

tivit

yof

the

det

ecto

r.It

isass

um

edth

at

secu

lar

equ

ilib

riu

mis

bro

ken

bet

wee

n230T

han

d226R

ain

the

238U

dec

ay

chain

.

Com

pon

ent

238U

lower

238U

upper

232T

h235U

[mB

q/kg]

Met

hyl

met

hac

ryla

tem

onom

er(L

Gb

on

din

g)

1.4

±1.

0<

15

<0.

9<

1.8

AV

acry

lic

<0.1

<2.2

<0.

5<

0.2

Acr

yli

cb

ead

s(R

PT

)<

3.1

16±

15

0.8

±0.3

0.6

±0.

5L

Gac

ryli

c<

0.1

<9.0

<0.

3<

0.6

304

wel

ded

stai

nle

ssst

eel

(ste

elsh

ell)

1.4±

1.1

<5.0

4.7±

1.5

<3.3

304

stai

nle

ssst

eel

stock

(ste

elsh

ell)

2.1±

1.1

<112

1.9±

1.1

<5.4

316

stai

nle

ssst

eel

bol

ts(s

teel

shel

l)<

6.1

<315

94±

9<

17

Car

bon

stee

l(s

tock

)2.0±

0.7

111±

43

10.0±

1.0

8.6±

1.9

Inva

rst

eel

(nec

k)

4.5±

1.5

120±

77

2.5±

1.5

<3.6

R59

12H

QE

PM

Tgl

ass

921±

34

225±

114

139±

725±

3R

5912

HQ

EP

MT

cera

mic

978±

56

15500±

2800

245±

28

503±

51

R59

12H

QE

PM

Tfe

edth

rou

ghp

iece

s1140±

60

2350±

1460

430±

32

38±

9R

5912

HQ

EP

MT

met

alco

mp

onen

ts<

5.5

−<

3.3

−R

G59

PM

Tca

ble

(Bel

den

E82

241)

4.5±

1.3

91±

46

1.2±

0.9

3.4±

1.4

PM

Tm

ount

PV

C(H

arvel

)72±

5232±

130

18.6

±2.

55.6±

1.5

PM

Tm

ount

cop

per

<0.

5<

10

<0.8

<1.

3N

eck

Vet

oP

MT

glas

s1230±

620

−407±

203

57±

29

Fil

ler

blo

ckp

olyet

hyle

ne

0.4±

0.3

<14

<0.1

<0.

15

Fil

ler

blo

ckS

tyro

foam

[40]

33.5

±3.

4115±

64

<1.5

<1.

4W

hit

eT

yve

kp

aper

(diff

use

reflec

tor)

<0.

350±

37

1.3±

0.8

<2.

2B

lack

Tyve

kp

aper

(LG

wra

pp

ing)

<1.

8<

127

5.6±

2.3

<3.

8B

lack

pol

yet

hyle

ne

tub

e(u

pp

ern

eck)

13.

7±

1.8

<60

3.2±

1.1

2.6±

1.4

TP

B(S

igm

aA

ldri

ch)

<3.

9−

<8.

7−

16

Tab

le7:

Gam

ma

ass

ay

resu

lts

for

toolin

gu

sed

du

rin

gco

nst

ruct

ion

an

dm

anu

fact

ure

of

det

ecto

rco

mp

on

ents

.A

ctiv

itie

sare

rep

ort

edw

ith

1-s

igm

au

nce

rtain

ties

.A

90%

con

fid

ence

lim

itis

pla

ced

wh

enth

em

easu

rem

ent

isb

elow

the

back

gro

un

dse

nsi

tivit

yof

the

det

ecto

r.It

isass

um

edth

at

secu

lar

equ

ilib

riu

mis

bro

ken

bet

wee

n230T

han

d226R

ain

the

238U

dec

ay

chain

.

Com

pon

ent

238U

lower

238U

upper

232T

h235U

[mB

q/k

g]

Pu

rifi

cati

on

Sys

tem

Wel

din

gT

IGw

eld

sam

ple

7.7±

5.7

<27

25.

2±

7.8

<16

SM

AW

wel

dsa

mp

le<

23

<1255

51.

9±

12.2

<13

Wel

din

gel

ectr

od

esA

(Blu

eD

emon

TE

2C

-116-1

0T

)221±

65

<493

1890±

184

<56

Wel

din

gel

ectr

od

esB

(Blu

eD

emon

TE

2C

-116-1

0T

)66.

6±

42.

6<

1300

710±

103

<138

Wel

din

gel

ectr

od

esC

(Blu

eD

emon

TE

2C

-116-1

0T

)86.1

±21.

8<

642

911±

73

<108

Wel

dfi

ller

rod

s<

4.8

<157

3.0±

2.5

<1.

8

Inn

erA

VS

an

din

gB

raze

dd

iam

ond

san

din

gp

ad(S

up

erab

rasi

ves)

141±

24

<845

49.

8±

17.

931±

19

Pla

ted

dia

mon

dsa

nd

ing

pad

(Su

per

ab

rasi

ves)

4680±

283

<4130

6180±

300

218±

64

3M60

02J

flex

ible

dia

mon

dp

ads

25.

1±

15.

4<

785

<10.8

<33

Dia

mon

dsa

nd

pap

er(D

iam

ante

Ital

ia)

3120±

136

<2300

3370±

125

157±

22

Red

san

dp

aper

(RP

T)

48.7

±19.

7<

335

<10.1

<32

LG

Acr

ylic

Poli

shin

gD

iam

ond

lap

pin

gfi

lm(3

M66

1X)

142±

38

<882

93.

6±

35.

0<

31

Dia

mon

dla

pp

ing

film

(3M

661X

)94.0

±16.

5<

276

105.±

18.1

<33

17

Table 8: Measured 222Rn emanation rates for components used in the DEAP-3600 detector contained within the stainless steel shell. A description of themost important components can be found in Section 4. Uncertainties arise fromcounting, sample emanation times, and detection and trapping efficiency.

Source Emanation Rate

[mBq/m2]Filler blocks 1.6 ± 0.5FINEMET PMT magnetic shielding [38] 0.8 ± 0.2ESR film reflector 4 < 2.2Tyvek diffuse reflector < 0.1Black tyvek absorber 0.4 ± 0.2PMT mount PVC (McMaster-Carr stock) < 0.7PMT polyethylene foam < 0.9Teflon sheets (McMaster-Carr stock) 0.4 ± 0.2High density polyethylene pipe 3.5 ± 0.8304 Stainless Steel (McMaster-Carr stock) < 1.6Carbon steel (McMaster-Carr stock) 0.6 ± 0.1White PMT mount adhesive styrofoam sheet < 1.5Stycast 1266 A/B (Emerson & Cuming) < 4.2

[mBq/m]RG59 PMT cable (Belden E82241) 0.026 ± 0.001Steel shell EPDM O-ring 16.1 ± 1.8Viton O-ring 1.3 ± 0.2Buna 451 O-ring 17 ± 2

[mBq/unit]Hamamatsu R5912 PMTs < 0.3PMT mount O-ring 0.3 ± 0.1

based on vaporization and subsequent chemical processing originally developed

by the Sudbury Neutrino Observatory (SNO) [43] was extended to allow a sensi-

tive assay of 210Pb in acrylic. Samples of the acrylic were vaporized, the residue

extracted by rinsing with an acidic Aqua Regia solution, and the effluent col-

lected and counted in the germanium well detector at SNOLAB. Details of the

DEAP-3600 acrylic vaporization process can be found in [43]. In addition, the

5.3 MeV alphas from the decay of 210Po were counted from the acid solution

by plating it out on nickel discs, with a 450 mm2 ORTEC ULTRA-AS ion-

implanted-silicon detector [43]. From these measurements, which affects the

18

210Pb surface alpha background, an upper limit of 0.62 mBq/kg 210Pb was set

for the AV acrylic.

Based on results from the extensive material assay campaign shown in Ta-

bles 6, 7, and 8, the expected number of background events in a 3-tonne-year

fiducial exposure is assessed. The expected number of alpha background events

assuming a 10 cm position resolution is shown in Table 9. The sensitivity of

the TPB assay given the amount of product available was not sufficient to test

adherence to the required alpha background. A number of steps were under-

taken during the synthesis of the TPB in coordination with the manufacturer

and during storage and deposition to control possible contaminations. The ex-

pected number of background events from neutrons produced in (α, n) reactions

and spontaneous fission is shown in Table 10, whereas neutrons produced from

(α, n) reactions and spontaneous fission due from radon plate-out on cold inner-

detector surfaces is shown in Table 11.

Table 9: Expected number of alpha background events based on screening mea-surements. Event numbers are given for the WIMP energy ROI with a 3-tonne-year fiducial exposure assuming a 10 cm position resolution.

Component 238U 232Th 210Pb

TPB < 6.8 < 9.9 −AV acrylic bulk < 0.3 < 1.3 < 0.3

Total Events: < 18.7 < 7.2 < 11.2 < 0.3

2.3. Material Production and Quality Assurance

Acrylic, or poly(methyl methacrylate) (PMMA), is a polymer of methyl

methacrylate (MMA). Alpha decays on the inner surface of the acrylic can pro-

duce nuclear recoils in the LAr, while those that decay in the bulk acrylic may

undergo the 13C(α, n)16O reaction [44] to produce neutrons, which can then

scatter in the LAr to produce a nuclear recoil. The acrylic annealing process

employed during construction to improve light collection also contributes to the

background from radon daughters due to the temperature-dependence of radon

19

Tab

le10:

Exp

ecte

dnu

mb

erof

neu

tron

back

gro

un

dev

ents

from

(α,n

)re

act

ion

san

dsp

onta

neo

us

fiss

ion

base

don

scre

enin

gm

easu

re-

men

ts.

Even

tnu

mb

ers

are

giv

enfo

rth

eW

IMP

ener

gy

RO

Iw

ith

a3-t

onn

e-yea

rfi

du

cial

exp

osu

reass

um

ing

a10

cmp

osi

tion

reso

luti

on

.

Com

pon

ent

238U

235U

232T

h

PM

Tgl

ass

(11.

1±

0.4

)·1

0−2

(3.5±

0.4)·1

0−3

(2.9

±0.2

)·1

0−2

PM

Tce

ram

ic(2.8

±0.

2)·1

0−3

-(2.6

±0.3

)·1

0−3

AV

acry

lic

<5·1

0−2

<2·1

0−2

<6·1

0−2

LG

acry

lic

<8·1

0−3

<3·1

0−2

<2·1

0−2

Fil

ler

blo

cks

(pol

yet

hyle

ne)

(1.1

±0.8

)·1

0−2

<3·1

0−3

<2·1

0−3

PM

Tm

ount

(PV

C)

(5.8

±0.4

)·1

0−3

(8.0±

2.0)·1

0−4

(3.8

±0.5

)·1

0−3

Nec

kst

eel

(1.1

±0.6

)·1

0−3

<6·1

0−4

(1.0

±0.6

)·1

0−3

Nec

kV

eto

PM

Tgl

ass

(5±

3)·1

0−4

-(4

±2)·1

0−4

Tot

alE

vents

:0.

230.

08

0.06

0.09

20

Table 11: Expected number of neutron background events based on radon ema-nation measurement assuming (α, n) reactions and spontaneous fission inducedfrom radon plate-out on cold inner-detector surfaces. Event numbers are givenfor the WIMP energy ROI with a 3-tonne-year fiducial exposure assuming a10 cm position resolution.

Radon Emanation Source Events

PMT cables (5.5 ± 0.2) · 10−3

AV acrylic < 5 · 10−4

LG acrylic < 3 · 10−3

Filler blocks (2.3 ± 0.7) · 10−2

PMT mount (PVC) < 4 · 10−3

FINEMET PMT magnetic shielding [38] (4 ± 1) · 10−3

Stainless steel shell < 6 · 10−3

PMTs < 8 · 10−3

Total Events 0.053

diffusion in acrylic. These acrylic-related alpha and neutron backgrounds are

mitigated by limiting radon exposure and controlling cleanliness during con-

struction.

2.3.1. Cryostat Acrylic

Reynolds Polymer Technologies, Inc. (RPT) Asia Ltd. (Rayong, Thailand)

acquired the MMA from the Thai MMA Co. plant in eastern Thailand [45]. For

DEAP-3600, the raw MMA was purchased directly from the production pipeline

and delivered to RPT Asia, limiting excessive exposure to radon-laden air in

storage tanks.

To calculate the radon load during manufacture the contamination when

MMA enters a production vessel was derived, assuming all radon in the volume

was dissolved in MMA and any radon progeny in the volume also becomes

trapped. The resulting 210Pb concentration in acrylic, using a density of MMA

of approximately 1000 kg/m3, is Aacrylic(210Pb):

Aacrylic(210Pb)[mBq/tonne] =λ210Pb

λ222RnAair(

222Rn)[mBq/m3]

where λ is the decay constant for 210Pb or 222Rn and Aair is the activity of 222Rn

in the environmental air. The radon levels at the MMA production plant were

21

measured with a Durridge Rad-7 electronic radon detector [46] to be consistent

with sea-level concentrations of 1 Bq/m3. The solubility of radon in solid acrylic,

8.2 [47], is used to estimate background levels after the panels have been formed.

Following all steps in production, the predicted 210Pb activity in the acrylic

panels of the AV is 0.021 mBq/kg [45], which is negligible compared to the

measured 210Pb upper limit in Section 2.2.

2.3.2. Light Guide Acrylic

The radiopurity requirements are less stringent for the LG acrylic as it is

not in direct contact with the LAr. Selection of the LG acrylic was based

on minimizing light attenuation. A wavelength-dependent attenuation length

was calculated from transmission measurements of 10 different length samples

ranging from 4 mm to 110 mm in a PerkinElmer Lambda 35 UV/Vis optical

spectrometer. Acrylic with a UV absorbing additive was used for both the

LGs and AV to minimize contributions from Cherenkov light generation in the

acrylic.

The attenuation length in acrylic can reach a few tens of meters for blue

light. After receiving and qualifying the LG acrylic material from the supplier,

more Rayleigh scattering was observed than had been present in earlier test

samples. This additional Rayleigh scattering could be reduced by annealing the

LGs near 85C. This procedure increased the transmission at 440 nm, the mean

TPB emission wavelength, by approximately a factor of two. Out of 10 acrylic

suppliers, Spartech Polycast acrylic, with an attenuation length at 440 nm after

annealing of 6.2 ± 0.6 m was selected for the LGs. The final acrylic delivered

by Reynolds Polymer for the AV had an attenuation greater than 7 m above

420 nm.

2.3.3. TPB Wavelength Shifter

The TPB required to cover the 9 m2 inner surface of the AV must be ra-

diopure to prevent alpha decays from the bulk of the TPB layer from producing

background events in the region of interest. For this reason, special arrange-

ments were made with the manufacturer, Alfa Aesar (Heysham, UK), in the

22

synthesis of the TPB to meet the required radiopurity requirements shown in

Table 3. Base chemicals with assay certificates of 99% purity or greater were

used during production, and all steps in the synthesis process were performed

under a boil-off nitrogen atmosphere. After production, the final product was

stored in a sealed vessel preventing exposure to humidity and UV light until

deposition in the AV.

2.3.4. Argon Purification System Components

The argon purification system components can contribute to the backgrounds

in the detector by mixing radon, deposited on or emanated from component sur-

faces, or other contaminants into the LAr target during filling or recirculation.

The sensitive surface area of the purification loop between the radon trap and

the detector is approximately 0.6 m2, which includes approximately 50 welds.

To mitigate potential contamination in the process tubing, argon-wetted

components of the process system were constructed with electropolished stain-

less steel and, where tooling was necessary, controlled TIG welding was per-

formed with gamma-assayed and certified ceriated welding stock. Physical

cleaning of stainless steel surfaces to remove dust and surface contamination,

ultrasonic cleaning cycles with Alconox precision cleaner and ultra-pure wa-

ter, and a chemical surface-layer contamination removal from the stainless steel

and weld material with citric acid passivation were performed, followed by a

final ultra-pure water rinse. After final assembly underground, the purification

loop was again passivated to remove contaminants plated-out onto the surface

from air exposure, and sealed until argon was introduced into the system for

purification.

3. Cryogenic System

The cryogenic system consists of a LN2 cooling system and a LAr purification

loop. An electropolished stainless steel cooling coil, shown in Figure 1, is filled

with LN2 to provide the necessary cooling power to condense and maintain the

23

argon in the detector in a liquid state between a temperature of 84–87 K and a

pressure of 13–15 psia.

The LN2 is gravity-fed to the cooling coil inlet through vacuum-jacketed

piping from a 3750-L storage dewar located above the detector in the Cube Hall

staging area, shown in Figure 24. Boil-off nitrogen gas is returned to the dewar

where it is re-condensed by three 1-kW Stirling Cryogenics SPC-1 cryocool-

ers [48]. During operation, two cryocoolers are operated continuously, with the

third available for backup or individual shutdown during routine maintenance.

The injection of lab-temperature argon from the purification system into

the AV during the initial cool-down of the AV created a high heat load on the

cooling system in addition to the load due to natural boil-off from the LAr in the

AV. The use of phase separators aided in preventing counter-current gas flow

and vapor lock throughout the cooling system and helped maintain a controlled

and stable cooling rate.

Temperature and pressure readings are taken throughout the cryogenic de-

livery system, and logged using an Emerson DeltaV slow controls system [49].

Valve automation, emergency shut-down, and isolation of critical components

can additionally be performed remotely through this DeltaV system.

Detector Cooling Coil. The cooling coil helix is 84 inches long and made of

0.75-inch outer diameter stock-size electropolished stainless steel tubing in a

5.50-inch center diameter helix with 35 turns at a 2.25-inch pitch. It is designed

to provide up to 1000 W of cooling, even with fully submersed in LAr. The LN2

is delivered to the bottom of the coil through a straight vacuum-jacketed supply

line in the center of the helix to prevent boiling gas from flowing up the inlet

line, counter-current to the downward liquid flow. The bottom of the straight

LN2 supply line curls upwards transitioning into the return helix, creating a

forced convective two-phase flow heat transfer.

At the bottom of the cooling coil, flow guides were designed to promote

convective LAr flow and block photons from scintillation events generated in

the neck region from entering the inner AV region. As the cooling coil was

24

designed to operate fully submerged in LAr, a detailed Computational Fluid

Dynamics (CFD) analysis was performed to optimize the geometry of the flow

guide, shown in Figure 2.

In the final running configuration of DEAP-3600, due to a seal failure in the

neck described in Section 3.2, a standing column of LAr is not maintained in

the neck. Instead, the LAr volume is maintained in the AV with an argon gas

volume in contact with the cooling coil.

3.1. Purification System

The design goal of the purification systems is to purify the argon target

to sub-ppb levels of electronegative impurities (CH4, CO, CO2, H2, H2O, N2

and O2) and to reduce the radon activity to as low as possible, nearing 5 µBq.

All argon-wetted components downstream of the cryogenic radon trap were

constructed from electropolished stainless steel or acrylic. For active compo-

nents, such as transfer pumps and purification components, certified use of

non-thoriated welding rods was demanded from the manufacturer.

The main purification system components, shown in a block diagram in

Figure 3, consist of the process pump, SAES getter, and custom-built radon

trap, condenser column, and boiler. The system was designed for a maximum

nominal flow rate of 4.9 g/s of argon and accepts argon gas at 300 K from a

bulk liquid storage tank.

Gas is injected into a KNF Neuberger 150.1.2.12 double diaphragm process

pump, which maintains a forward pressure of 30 psig at the top of the system.

The double diaphragm model safely imposes an extra barrier between the lab

air and process gas in the event of pump failure, and the pressure between the

two diaphragms is monitored.

Chemical purification is performed by a SAES Mega-Torr PS5-MGT15 hot

metal getter custom fabricated for DEAP to avoid internal components with

thoriated welds. It is specified to accept 99.999% high-purity argon at a maxi-

mum flow rate of 7.4 g/s. The getter contains a safety interlock system which

prevents over-heating and ignition of the getter material in the attempt to purify

25

Process Flange

Cryogenic

Seal

Flow Guide

Assembly

Acrylic

Vessel

Neck Vacuum

Jacket

Butyl O-rings

Acrylic

Neck

Inner flow

guide detail

Outer flow

guide detail

Figure 2: Left: A schematic of the internal neck. The vapour space outside thevacuum-jacketed neck (orange) and inside the outer steel neck (green) beganto fill with LAr during the first AV fill, causing the failure of the butyl O-ringseals between the acrylic and steel neck interface. Right: Concentric inner andouter acrylic flow guides in AV neck designed to guide the convective liquid flowpattern. The flow guides were assembled from a stack of machined and sandedacrylic discs. A piston ring (pink) at the bottom of the flow guides covers thegap between the outer flow guide and inner acrylic neck.

gas that is too high in electronegative impurities.

Radon and radioactive impurities are removed by absorption in a custom-

built charcoal trap, designed to take gas at 300 K from the getter, pre-cool it

to 100 K, and pass it through a charcoal column. For optimal performance, the

26

Argon

Dewar

OutletFlow

Control

Process

Pump

SAES GetterRadon TrapCondenser

Argon

Dewar

Inlet

Boiler

To DEAP-3600

Vessel

From

DEAP-3600

Vessel

Figure 3: Flow diagram of the DEAP-3600 purification system. Gas is injectedinto the loop ahead of a flow controller and process pump which circulates theargon through a SAES getter, radon trap, and condenser before entering theDEAP-3600 vessel. Argon returns from the vessel through a boiler unit to com-plete the loop.

trap should be as cold as possible while maintaining argon in the gas phase.

The trap is placed between the active purification system components and the

detector to minimize emanated radon from the system itself from mixing into

the argon volume. The inlet is surrounded by a copper block, partially immersed

in LN2, with tunable cartridge heaters, capable of 600 W of heating, to prevent

argon gas from freezing. The charcoal cartridge is a 12-inch cylinder with a

3-inch diameter filled with 610 g of Saratech charcoal, selected to have very

low radon emanation. The charcoal is contained by a barrier of stainless steel

wool, retaining steel mesh, and 50-µm VCR filter gaskets on both the top and

bottom to prevent particulates from escaping. The cartridge is surrounded by

a bake-out heater. All components of the radon trap are contained within an 8-

inch-diameter cylinder, wrapped in multi-layer insulating foil, and housed within

a 10 inch vacuum space.

The custom-built condenser column is built to liquify gas from the radon

trap before delivering it into the detector. The condenser comprises a stainless

27

steel coil formed from a 39-ft long, 0.5-inch-outer diameter stainless steel tube,

suspended inside an 8-inch-diameter cylinder and immersed in LN2. Either

liquid or gaseous argon can be delivered to the AV; for gaseous argon it is

liquified on the neck cooling coil before dripping into the AV.

Purified argon gas is directed into the AV via the inlet on the main process

flange. The system was designed so that liquid could be extracted and then

vaporized in a boiler before being returned to the purification system. In the

current configuration, with gas in the detector neck, gas is returned to the

boiler inlet. An auxiliary gas return, running in parallel to the main outlet also

connects the process flange to the purification loop before the boiler unit. To

avoid direct exposure of the heating elements to the argon, a 1.5-kW heater is

coiled around the stainless steel flow line returning from the AV to vaporize the

LAr. Gas exiting the boiler is delivered back through the flow control valve into

the KNF pump at the the top of the loop.

If any of the relief valves throughout the system opens and fails to close fully,

it is possible to back-stream lab air into the purification system. To mitigate this

potential contamination, a double check valve assembly is used on all pressure

relief assemblies, shown in Figure 4. A 3 psig check valve is in contact with

the argon gas. Behind this inner check valve is a small enclosed volume with a

valve to purge with argon gas, pressure gauge to indicate a pressure relief, and

a variable pressure outer relief valve to provide a redundant seal.

The process systems may be operated in 3 main configurations: filling, recir-

culation, and storage recirculation. After the initial purification of argon when

filling the detector, after one year of running, re-purification of the argon has

not be necessary.

Filling — During filling, gas is constantly drawn from storage by the purifi-

cation system, purified, and injected into the detector. The gas can be liquified

before delivery into the AV or can be injected as gas and liquified by the cooling

coil. The injection rate can be varied between 0–4.9 g/s depending upon the

cooling rate of the AV and the available cooling capacity.

Recirculation — After filling, or during pauses, the system may be oper-

28

Figure 4: Double check valve relief assembly used throughout the purificationsystem. A check valve in direct contact with the purification system leads tosmall volume back-filled with clean argon gas, a pressure gauge, purge valve,and variable pressure relief valve.

ated to recirculate boil-off gas around the process loop, re-purifying and liquify-

ing in a steady state. If the detector is full, liquid may be extracted directly,

boiled, and recirculated around the loop.

Storage recirculation — The detector may be bypassed and cold purified

gas extracted after the radon trap and transferred back to the 3750-L storage

dewar. This is the default mode of the system, allowing for stabilization of the

purification system during start-up and pre-purification of the gas space in the

storage dewar.

3.2. Neck Seal Incident

A vapor space exists between the outer vacuum-jacketed neck (shown in

orange in Figure 2) and the inner surface of the stainless steel neck (green).

There is a small fit-tolerance between the vacuum-jacketed neck and the inner

surface of the acrylic neck (grey). The piston ring (described and shown in pink

29

in Figure 2) does not produce a true seal at the bottom of the neck to prevent

LAr from filling into the vapor space.

During the initial filling of the detector, the LAr level rose through the acrylic

neck and flow guides. A leak at the connection between the process flange (blue

in Figure 2) and inner vacuum-jacketed neck prevented the hydrostatic head

pressure needed to keep LAr from filling the outer space between the vacuum-

jacketed neck and acrylic neck. The rising LAr filled equally the inner space

contained by the flow-guide assembly and the space between the outer vacuum-

jacket and inner acrylic neck, coming in direct contact with the acrylic.

The acrylic to steel neck interface is sealed with 2 butyl O-rings and an

additional cryogenic seal, shown in Figure 2, designed to contract and seal when

slowly cooled. The rapid temperature drop due to direct exposure to LAr in

the neck acrylic lead to failure of the butyl seals, allowing a pathway for clean,

radon-scrubbed boil-off nitrogen that was purging the steel shell volume to leak

into the AV volume. On 17 August 2016, a contamination of approximately

100 ppm N2 leaked into the LAr causing a sharp decrease in the observed long

time constant to argon scintillation and a spike in the AV pressure. This required

complete venting and boiling of the approximately 3600 kg LAr in the AV.

After the AV was emptied of all cryogen and the cause of the seal failure

identified, clean Ar gas was injected and liquified in the AV. A reduced fill level

was chosen to minimize the possibility of LAr reaching the butyl seals again. A

final fill level corresponding to approximately 3300 kg LAr was chosen and has

been stably maintained since completing the second fill in November 2016.

4. Inner Detector Construction

The inner detector consists of all elements between the LAr and the steel

shell, as summarized in Figure 5. The construction of these components is

described in this section, with the exception of the PMTs (see Section 5.1).

Acrylic Vessel. Cast acrylic is mechanically strong enough that few metal sup-

ports are required to hold the weight of the AV, LAr, LGs, filler blocks, and

30

Polyurethane Foam Insulation

Filler Block Styrofoam SM Insulation

Filler Block High-Density Polyethylene

Stainless Steel Shell

Acrylic Vessel

R5912 PMT

PVC PMT Mount

Copper Thermal Short

Stainless Steel Mesh Acrylic Light Guide

Light Guide Stub

PMT Mount Springs

Diffuse Reflector

FINEMET Magnetic Shield and Specular Reflector

Figure 5: Cross section of the inner detector components from the acrylic vesselto the stainless steel pressure vessel.

PMTs. The AV was built in three pieces (the neck, collar, and the truncated

sphere) due to the limited envelope of the mine shaft leading underground.

The average vessel radius is estimated at 846 mm when cold and 7 mm larger at

room temperature. The neck provides mechanical support of the AV from above

and access for the purification system cooling coil. The inner-neck diameter is

255 mm.

Each of these three acrylic pieces underwent a different construction:

• The neck raw material was made from acrylic sheets bonded together to

form a rough cuboid.

• The collar was made of a single, thermoformed panel of acrylic.

• The truncated sphere consisted of 5 spherical slices and one polar cap,

thermoformed and then bonded together by RPT.

The acrylic sphere was machined to the final spherical shape, shown in Fig-

ure 6, on a 5-axis computer numerical controlled (CNC) mill at the University

of Alberta. LG stubs were then machined to receive the LGs. Particular care

31

was given to avoid unnecessary exposure of the acrylic to any material that

might generate surface stress, such as methanol and MMA. The machining was

performed using the guidelines from Stachiw [50].

Figure 6: AV as delivered by Reynolds Polymer Technologies (left) and aftermachining (right) at the University of Alberta.

The AV sphere, collar, and neck were then shipped to the SNOLAB un-

derground facility and bonded together by RPT. Due to the large tolerances

produced in the bonding process, additional post-machining underground was

required.

Flow Guides. The acrylic used to construct the AV was also used for the flow

guide assembly, shown in Figure 2. These were milled at the University of

Alberta in a controlled room with a radon level of 0.3 mBq/m3. Once machined

to shape, approximately 20 µm was removed by hand-sanding in a nitrogen-

purged glovebox at Queen’s University with an oxygen content below 20 ppm.

Light Guides and PMT Mount Assembly. Each of the 255 light guides is 45 cm

long and 19 cm in diameter, giving an AV surface coverage of 76%. The outer

LG face is concave to match the geometry of the PMT face. The annealing,

machining on CNC lathe, and polishing of the LGs were completed at TRIUMF

in British Columbia, Canada.

The LGs were bonded on the AV underground using a tripod set in position

with an alignment cylinder suctioned to the target stub, and attached to the

AV by clamping to neighboring stubs or LGs, shown in Figure 7. The LG to be

32

bonded was constrained in the tripod to only move perpendicularly to the stub

face with a compression spring.

Figure 7: A light guide (right) ready to be bonded on the AV (left).

The bond was created by constructing a form-fit polyethylene dam with a

radial bulge around the outer perimeter of the LG-stub gap interface. This

bulge allowed for an excess reservoir of acrylic monomer to prevent concavity

in the bond during polymerization, and was machined off after a post-bonding

anneal. The dam was sealed against the stub with O-ring clamps. A computer-

controlled fill system was used to ensure consistent bonds.

Although approximately 80% of photons emitted in the LAr are trapped

in the LGs by total internal reflection, an additional specular reflector, loosely

wrapped around the LGs, is used to increase light collection and provide optical

isolation between LGs. Aluminized mylar was selected for its high reflectance

and lack of alpha-induced scintillation. A 50-µm mylar foil was sputter-coated

by Astral Technology Unlimited (Northfield, MN, USA) with 100 nm of alu-

minum, using a 99.999% high purity custom-made sputtering target by Laurand

Associates (Great Neck, NY, USA) to satisfy the radiopurity requirements.

Areas on the spherical AV shell between LGs were covered in a diffuse, 98%

reflective white Tyvek base layer followed by a layer of black Tyvek and a layer

of closed-cell polyethylene foam backing. These layers help to maximize the

33

light collection and keep stray photons originating outside the active volume

from leaking into the LAr volume.

At the end of each LG, a PMT mount assembly, detailed in an exploded

view in Figure 8, supports and maintains optical coupling between the PMT

and LG. A cylindrical PVC barrel seals to the end of the LG with an O-ring and

a lightly-compressed backing yoke. The LG–PMT interface volume is filled with

silicone oil (Sigma Aldrich #378399) serving as an optical coupling, which was

found to be a good index match between acrylic and glass, and has a favorable

coefficient of thermal expansion and viscosity. The assembly is wrapped in a

sleeve of FINEMET [38] magnetic shielding (Section 5.1) and an outer sleeve

made of copper acting as a thermal short to passively warm and prevent thermal

gradients across each PMT.

PMT base housing

PMTDouble O-ring sealOil fill portsRetaining springs (3)Light guide

Band clampPVC PMT mount

Copper thermal short

Magnetic shieldingsleeve

Figure 8: Exploded view of PMT mount assembly and light guide.

Filler Blocks. The volume between LGs is filled with 486 filler block stacks fab-

ricated from alternating layers of high-density polyethylene and Styrofoam [40].

The polyethylene and Styrofoam combination provides superior thermal insu-

lation and equivalent neutron shielding compared to acrylic alone. Retaining

springs at the warm end of the LG push the filler blocks against the AV and

keep the blocks centered between LGs when the acrylic contracts during cool-

34

down. A 5 mm gap between the LGs and filler blocks ensure adequate spacing

for thermal expansion without generating stress.

Temperature Sensors, Foam Insulation, and Steel Mesh. The inner detector is

instrumented with an array of 85 PT-100 resistive temperature detectors. For

each selected filler block, a sensor was bonded to the bottom, middle and top

block layers, for temperature measurement at a distance of 0.9 m, 1.1 m, and

1.3 m, respectively, from the center of the AV. A fourth sensor was bonded to

the copper thermal short on a LG adjacent to the filler block.

The copper thermal shorts are surrounded by pieces of open cell polyurethane

foam for additional thermal insulation. A stainless steel mesh is fastened around

the entire detector to contain shattered components in the event of an AV struc-

tural failure that could block the vapor relief path along the steel neck. The

volume behind the mesh and inside the steel shell is maintained as a vapor

space, and is continuously purged with radon-scrubbed boil-off nitrogen gas.

Figure 9 shows the stages of installation during construction of the PMTs,

filler blocks, and foam insulation on the AV.

4.0.1. Acrylic Treatment

Annealing. The AV was annealed five times during its construction underground

at a temperature between 80C and 85C to relieve stress in the material after

bonding and to harden bonds. These anneals occurred (i) after machining of the

AV truncated sphere, shoulder, and neck; (ii) after bonding the AV truncated

sphere, shoulder, and neck together; (iii) after underground machining of the

neck stubs, neck and shoulder bond areas, and before LG bonding; (iv) after

LG bonding and before bond-bulge removal; and (v) after all bonding was com-

pleted, a final extra anneal. A 10-ft cubical oven was constructed from foam

insulation panels to perform the annealing. An external heater and blower cir-

culated heated air in the oven. Several smaller ducts were positioned across the

oven floor to distribute the reheated air. The difference in temperature across

the oven was required to be less than 2C for annealing.

35

Figure 9: A) The acrylic vessel after bonding on the light guides. B) Reflectorsand magnetic shielding installed around light guides. C) View from inside thedetector with the white Tyvek and most PMTs installed on light guides. D) De-tector with filler blocks installed and during PMT installation. E) Detector withall PMTs installed and during backing foam installation.

As radon diffusion increases with temperature, the level of radon in the air

inside the AV was monitored and controlled to below 10 Bq/m3 when annealing.

During the first three anneals, heated air that was piped underground from the

surface was blown into the vessel. This surface-air is lower in radon by approxi-

mately a factor of 10 with respect to the underground lab air. After bonding the

neck to the AV, the inside volume was flushed with lab-grade nitrogen (fourth

cycle) or argon (fifth cycle) and then kept sealed during annealing. Using the

exposure history and concentration measurements at each stage of construction

and annealing, the calculated build up of radon progeny from the integrated

radon exposure of the AV would result in a 210Pb activity of 14 mBq/kg in the

surface acrylic prior to resurfacing.

36

Acrylic Resurfacing. A robotic sander, the “resurfacer” shown in Figure 10, was

designed to remove up to 1 mm of the inner surface layer of the spherical AV.

Two rotating sanding heads with 3M Flexible Diamond QRS Cloth Sheet 6002J

M74 sanding pads were drawn across the inner surface of the AV to remove

and collect the sanded acrylic under a continuous flush with ultra-pure water.

When sanding, a linear actuator pushed the sanding motor outwards against

the acrylic, developing a constant, nominal 12-lb normal force against the AV.

In each sanding pass, the tilt arm moved in a spiral pattern from its deployed

vertical position down to the equator and back, with the opposing arm section

sweeping out the region between the south pole and equator. The starting

azimuthal angle was incremented by 60 before the start of the next pass to

provide a more uniform coverage across the sanding surface. After sanding of

the surface layer was completed, the sanding heads were retracted, and the

acrylic surface was flushed with ultra-pure water to extract remaining loose

material.

The resurfacer was deployed through the AV neck and operated with the

detector volume hermetically sealed from the lab and continuously purged with

radon-scrubbed boil-off nitrogen gas [51]. After sanding, the resurfacer was

extracted through the glovebox and into an auxiliary canister which was sealed

and purged with radon-scrubbed boil-off nitrogen, avoiding any exposure to

laboratory air.

The integrated sanding time over the inner AV surface was approximately

198 hours. Based on this sanding time and the measured acrylic removal effi-

ciency, an estimated 500± 50 µm of acrylic was removed from the inner surface

of the AV. This thickness is sufficient to reduce the 210Pb surface backgrounds

down to near the upper assay limit of 2.2 × 10−19g/g (0.62 mBq/kg), as shown

in Figure 11.

Wavelength Shifter Deposition. An 11-cm-diameter spherical evaporation source

was constructed from 316 stainless steel, shown in Figure 12, for the applica-

tion of the TPB wavelength shifter to the inner surface of the AV. Details on

37

Figure 10: Overview of the resurfacer device when deployed in the acrylic vessel.

the prototyping and testing of the source can be found in [52] and [53]. An

inner copper crucible, which holds the TPB powder, was radiatively heated by

a flexible Watlow 125CH93A1X coil heater wound on the outside of the evapo-

ration source. As the source was heated, the TPB molecules scattered inside the

source before exiting through one of the twenty 14-mm-diameter holes, creating

a uniform outgoing flux.

After the resurfacing of the AV, the empty TPB source was deployed to the

38

Figure 11: Calculated 210Pb alpha activity in the AV before resurfacing. Ac-tivity after radon-laden air exposure (9 months on surface, 6 months in mineair and 1 month in radon reduced air) is shown in blue. Activity due to radondiffusion into acrylic is shown in red. The cyan line is from acrylic assay upperlimit of 2.2× 10−19g/g in the AV. The 500 ± 50 µm of removed acrylic reducesthe activity of 210Pb down to near the assay upper limit.

center of the AV for a vacuum bake of the acrylic. The inner surface of the

AV was brought to 50C to outgas absorbed water and reach the approximately

10−6 mbar vacuum needed for the deposition. Two evaporations of 29.4 ± 0.2 g

combined total mass were performed on the AV, which yielded a uniform coating

thickness of 3.00 ± 0.02 µm. A detailed description of the TPB deposition can

be found in [9].

5. Light Detection Systems

5.1. Photomultiplier Tubes

Hamamatsu R5912 8-inch-diameter HQE PMTs [14] were selected for DEAP-

3600 for their high photon detection efficiency (nominal 32%), low dark noise

39

Figure 12: TPB evaporation source. A heating wire wraps the outside of thesphere, radiatively heating the copper crucible, which is contained within thesphere, evaporating the TPB powder.

rates, and good timing characteristics. Details on the PMT characterization can

be found in [54], while ensemble characterization versus PMT number (PMTID),

which is closely coupled to the vertical location of the PMT, for all 255 LAr

PMTs is presented herein.

The PMTs operate at bias voltages between 1500 V and 1900 V, described in

Section 6. In-situ measurement of the mean single photoelectron (SPE) charge

for all PMTs is shown in Figure 13 and has a mean of 9.39 pC and a RMS of

0.16 pC. Two outliers are PMTs that developed faults in the base. The mean

SPE charges are monitored on an ongoing basis and are related to the applied

bias voltage through:

q = A · Vγ

where q is the mean SPE charge, V is the bias voltage, and A a normalization

parameter. The γ parameter was measured for most PMTs5 and is also shown

5Data were not available for 36 PMTs as some were kept at their nominal voltage to verifythat the LED light intensity did not vary between runs, and some DAQ channels had not yetbeen configured.

40

in Figure 13 with a distribution mean of 6.9 with an RMS of 0.2.

PMTID0 50 100 150 200 250

Mea

n S

PE

cha

rge

[pC

]

7.58.08.59.09.5

10.010.511.0

Counts/bin10 20 30

7.5

8.08.5

9.0

9.5

10.0

10.511.0

PMTID0 50 100 150 200 250

γ

5.05.56.06.57.07.58.08.59.0

Counts/bin10 20 30

5.05.56.06.57.07.58.08.59.0

Figure 13: Top: Mean single photoelectron charge vs. PMTID with a mean of9.39 pC and an RMS of 0.16 pC. Bottom: γ parameter vs. PMTID, with a meanof 6.9 and RMS of 0.2, that relates the mean SPE charge to the bias voltage isshown vs. PMTID.

The dark noise rates per PMT, shown in Figure 14, are strongly temperature-

dependent and are therefore shown while the detector was at room-temperature

and right after cool-down when the PMTs were close to their operating temper-

ature, with those near the top of the detector at 280 ± 2 K and those near the

bottom at 260 ± 2 K. The data were taken while argon gas was inside the AV.

The room-temperature dark noise rate has a mean of 5.80 kHz and an RMS

of 0.78 kHz. The dark noise distribution with cold argon gas has a mean of

0.24 kHz and an RMS of 0.06 kHz. When the detector is filled with LAr and

the PMTs are at their operating temperature, the true dark noise rate cannot

be measured due to the high rate of 39Ar events.

The full-width at half-maximum (FWHM) transit time spread, due to the

variance in time from photoelectrons (PEs) liberated off the photocathode to

41

impingement on the first dynode, was measured ex-situ using a tagged 90Sr

source. The distribution, shown in Figure 14, has a mean of 2.60 ns across all

255 PMTs with an RMS of 0.12 ns.

PMTID0 50 100 150 200 250

Dar

k N

oise

rat

e [H

z]

210

310

410

T = 295 KT = 260-280 K

Counts/bin10 20 30

210

310

410

PMTID0 50 100 150 200 250T

rans

it tim

e sp

read

[ns]

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Counts/bin10 20

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Figure 14: Top: Dark noise vs. PMTID for room-temperature (295 K, pink)and just after filling the detector with argon (260 K for large PMTID near thebottom of the detector and 280 K for small PMTID near the top of the detector.The room-temperature dark noise distribution has a mean of 5.8 kHz and anRMS of 0.78 kHz. The cold dark noise distribution has a mean of 0.24 kHz andan RMS of 0.06 kHz. Bottom: Full width at half maximum transit time spreadvs. PMTID, with a mean of 2.6 ns and an RMS of 0.12 ns. Error bars shownare statistical and typically smaller than the marker size.

Afterpulsing occurs in distinct time regions between 100 ns and 10 µs and

is caused when residual gas inside the PMT becomes ionized by moving elec-

trons. The probability per PMT shown in Figure 15 is the total probability of

observing an afterpulse of any charge per SPE pulse within that time window.

The relatively high afterpulsing probability has an mean of 7.1% and an RMS

of 1.8%, which will be mitigated in analysis [54].

The double and late pulsing probabilities for the 255 PMTs are shown in

Figure 16. In a double pulse, the full SPE charge is split over two separate

42

PMTID0 50 100 150 200 250

Tot

al A

P p

roba

bilit

y [%

]

02468

101214

Counts/bin10 20 30

02468

1012

14

Figure 15: Total afterpulsing probabilities vs. PMTID. The distribution has amean of 7.1% and an RMS of 1.8%.

pulses due to inelastic scattering of the photoelectron on the dynode. A late

pulse is a similar effect, but in which the photoelectron backscatters from the

first dynode without producing any secondary electrons. The distribution of

double pulses has a mean of 2.7% and RMS of 0.2%, while the mean and RMS

of the late pulse distribution is 2.3% and 0.1%, respectively. Characterization

of these probabilities is used for time-resolved studies of pulse shapes at short

times and time-based event position reconstruction.

Magnetic Field Suppression. A combination of field-compensating coils around

the DEAP-3600 detector and an individual FINEMET shield around each PMT

is used to mitigate external magnetic fields which can affect PMT efficiency

and reduce gain. Although FINEMET saturates at a maximum flux density

of 1.13 T, a small 60 µm thickness has a maximum relative permeability of

70 kH/m [38], which makes it competitive with significantly thicker and heavier

sheets of mu-metal.

Four compensation coils, each 11 feet in diameter, attached to threaded rods

at ± 0.75 m and ± 2.25 m from the equator, are submerged in the water shield

tank (Section 7). This configuration was determined from a measurement of

the ambient field in the detector surroundings using a hand-held geomagne-

tometer (Integrity Design IDR-321) and modeling with Radia [55]. The coils

are operated with remotely-controlled switching mode 600 W power supplies

43

PMT ID0 50 100 150 200 250

Dou

ble

Pul

se P

roba

bilit

y [%

]

1.61.82.02.22.42.62.83.03.23.4

Counts/bin5 10

1.61.82.02.22.42.62.83.03.23.4

PMT ID0 50 100 150 200 250La

te P

ulse

Pro

babi

lity

[%]

1.61.82.02.22.42.62.83.03.23.4

Counts/bin5 10

1.61.82.02.22.42.62.83.03.23.4

Figure 16: Top: Distribution of double pulse probabilities for the 255 innerPMTs vs. PMTID with a mean of 2.7% and RMS of 0.2%. Bottom: Distributionof late pulse probabilities for the 255 inner PMTs vs. PMTID with a mean of2.3% and an RMS of 0.1%.

(Extech 382275). This arrangement leaves only an uncompensated 140 mG

radial component, which is further reduced with the FINEMET shielding wrap-

ping each PMT. Six 3-axis fluxgate magnetometers (Stefan Mayer Instruments

FLC3-30 supplied with waterproof enclosures) were installed on the outside of

the steel shell to monitor the field near the detector. These are read out through

the DeltaV slow control system and logged for diagnostic purposes.

5.2. Neck Veto System

The bottom 10 cm of the AV neck was wrapped with 100 Kuraray Y-

11 (200M) wavelength shifting optical fibres, shown in Figure 17, to provide ad-

ditional information in identifying and discriminating backgrounds (e.g. Chere-

knov light) from this region of the detector where there may be incomplete light

collection. Without full collection, a relatively high energy event could recon-

struct with a low number of detected photoelectrons, potentially leaking into

44

the energy ROI for the dark matter search. Each fibre is between 2.6 m and

3.3 m, and the fibre ends were gathered into bundles of 50 and optically cou-

pled to four Hamamatsu HQE extended green R7600-300 PMTs. The PMTs

were placed on top of the filler blocks near the neck at the same distance as the

primary PMTs from the LAr volume. The neck veto PMTs are read out by the

same DAQ hardware and software used for the inner detector PMTs.

Figure 17: The neck veto wavelength shifting fibers wrapping the bottom 10 cmof the AV neck before installation of the neck filler blocks. The fiber bundleswere then optically coupled to R7600-300 PMTs (not shown).

5.3. Calibration Systems

Calibration systems can be deployed in and around the detector to character-

ize the PMT response and event reconstruction. The optical calibration systems

are comprised of a diffusing laserball source deployed inside the detector before

it was cooled to characterize the optical response of detector materials, and

an array of permanently-installed fiber optic cables, coupled to LED-drivers,

to monitor PMT gains and pedestals as a function of time. All external ra-

dioactive calibration sources are deployed outside of the steel shell. A tagged

22Na gamma source is used to monitor the energy scale, resolution and position

response, and a tagged americium-beryllium (AmBe) neutron source is used

to monitor the response to neutron-induced nuclear recoils as expected from

45

WIMP interactions. An untagged 232U source (15.6 kBq in March 2017) was

deployed to study the detector response to gamma interactions in the acrylic

producing Cherenkov light. Additionally, the intrinsic radioactivity from 39Ar

beta decays provides a source of uniformly-distributed events with a known en-

ergy spectrum used to calibrate energy and position reconstruction biases from

50–5000 photoelectrons.

5.3.1. Optical Calibration

The initial optical calibration of the detector was performed with a nearly-

isotropic optical photon source deployed once into the inner detector volume

before the LAr fill. The laser-driven diffuse light source was based on the cali-

bration source from the SNO experiment [56]. A PerFluoroAlkoxy (PFA) flask,

11 cm in diameter, was loaded with 50-µm quartz beads suspended in Silicone

RTV-2 gel (Wacker Silgel 612 A/B) to produce a diffusing medium. Three laser

diode heads with wavelengths of 375 nm, 405 nm, and 445 nm injected light from

a Hamamatsu PLP-10 picosecond light pulser (70 ps typical FWHM) through

a 1-mm-diameter optical fibre (Mitsubishi Rayon Co., LTD. SH-4001) into an

acrylic stub with an end face centered in the PFA flask. These wavelengths

were chosen to be below, near, and above the excitation wavelength of TPB;

the highest wavelength has sensitivity only to the acrylic optical properties while

the lowest wavelength additionally has sensitivity to the TPB coating optical

response. The fast pulse additionally allowed for a measurement of the channel-

by-channel timing offsets. The source was deployed in the AV using the same

deployment system used for the TPB deposition to the center of the detector

(Z=0) and ± 55 cm, along with four 90 azimuthal rotations at each z-position

to disentangle non-uniformity in the source itself. The uniformity of the source

was additionally measured using a CCD camera, ex-situ, to be within ± 10%

with a 5% measurement uncertainty.

To calibrate the detector response as a function of time, optical calibration

is performed with a permanently-installed LED light injection system. There

are 22 light injection points in the detector, with 20 uniformly spread across

46

the PMT array, and 2 located in the neck region. Additionally, there is one

injection point in the water shield tank to calibrate the muon veto PMTs. At

each point for the PMT array, a cylindrical section made from acrylic and coated

in aluminum is attached to the LG perpendicular to a PMT face. Fast electronics

based on [57] drive 435 nm LEDs to generate the light injected into the fibres,

and an avalanche photodiode is used to monitor and correct for variations in

LED intensity on a run-by-run basis. Injected light travels through a 1-mm-

diameter acrylic fibre to the cylindrical section, where it is reflected onto the

PMT face. The LED-calibration PMT detects the light, and approximately 20%

is scattered down the LG into the inner detector volume, providing a diffuse

light source to all PMTs more than approximately 50 degrees away from the

active LED. The distribution of light from this calibration source can be seen

in Figure 18, where the occupancy is defined as the fraction of the total light

flashes that are registered in a given PMT.

1 PMT/bin (sorted by angular distance to active LED)0 50 100 150 200 250

PMT

occu

panc

y (%

)

0102030405060708090

100

Angle to active LED (deg)0 50 100 150

Figure 18: Observed PMT occupancies from an LED calibration run vs. PMTsorted by distance. The outer plot shows the average occupancy for each PMT,where the PMTs are sorted by ascending angular distance to the active LED.The inlay plot shows the PMT face positions projected onto a plane and coloredby their occupancy. The active LED is marked as ‘LED’ on the inlay plot. At anangular distance larger than 50 degrees from the active LED, the LED systemprovides a diffuse source of light.

A measurement of the the relative PMT channel efficiency was performed

47

using the diffuse laserball and LED calibration system. The relative efficiency is

defined with respect to a single PMT channel. A strong correlation between the

efficiency measured with each source is shown in Figure 19. The RMS spread

of the PMT channel efficiencies is measured to be 3.5%.

Figure 19: Correlation between relative PMT channel efficiency measured withthe laser optical calibration source vs. the relative efficiency measured with theLED calibration system. The blue line represents the perfect correlation sce-nario. The two independent measurements agree within 7%.

5.3.2. Radioactive Calibration

There are 3 vertical stainless steel tubes (A, B (not visible), and E in Fig-

ure 20) used to deploy the tagged gamma and neutron calibration sources around

the detector equator and one stainless steel tube (C in Figure 20) projecting out-

ward from the upper hemisphere. A circular high density polyethylene tube (F

in Figure 20) wrapping around the detector is also used with the gamma sources.

A pulley and carriage system is used to deploy the gamma and neutron source

canisters, and is driven by a Mclennan 34HSX-208E stepper motor through a

Mclennan SimStep controller. The deployable source position uncertainty was

measured ex-situ to be approximately 1 cm.

48

Figure 20: The DEAP-3600 detector in the water shield tank with the verticalcalibration tubes A, B (not visible; in background), C, E, and circular highdensity polyethylene tube, F, indicated. The 48 PMTs attached to the steelshell are used for the muon veto.

The 1 MBq 22Na source (created April 2012) is contained between two

Cerium-doped Lutetium Yttrium Orthosilicate (LYSO, Hilger [58]) crystals to

tag the back-to-back 511 keV annihilation photons. The 20-mm-diameter, 20-

mm-long LYSO crystals are read out with two compact Hamamatsu R9880U

PMTs [59]. The tagging system can be compact, as 50% of the 511 keV photons

are attenuated in 8.5 mm in LYSO.

The 74 MBq AmBe neutron source is surrounded by two Ametek-packaged

NaI crystal and PMT assemblies. Each 40-mm-diameter, 51-mm-thick NaI crys-

tal has a 10 mm by 10 mm well to contain the radioactive source, and is coupled

to a 38 mm ETL 9102 PMT with an internal Cockcroft-Walton high voltage

49

generator allowing the PMT voltage to be driven from a 5 V source.

Neutrons are created when an alpha produced by the decay of 241Am cap-

tures on a 9Be nucleus releasing a neutron and a 4.4 MeV gamma. The tagging

PMT thresholds are set such that the gamma is used as a tag for the emitted

neutron. The AmBe source is wrapped in 2 mm of lead foil to remove 99.9% of

the 60 keV gammas produced in addition to the alphas from the decay 241Am.

The neutron rate from the source, as measured by the manufacturer (Eckert

and Ziegler Isotope Products), is 4.8 kHz.

6. Electronics

The overall architecture of the data acquisition (DAQ) system is shown in

Figure 21. PMT signals are analyzed by the Digitizer and Trigger Module

(DTM), which decides whether to trigger event readout. Trigger signals are sent

to commercial digitizers (CAEN V1720s [60] and V1740s [61]), which digitize

the PMT information. The digitized information is then read out, filtered, and

written to disk.

MPOD HV

Trigger signals

ASUMs

Event builder and filter

Disk

Test pulses

deap00

slow

fast

Light injection

Input Signal processing Software

slow

Clock signals

Readout

Readout & filter

Readout & filter

Readout

deap05

deap01..04

lxdeap01

deap05

High voltage

4 SCBs

22 SCBs 255 LAr

PMTs

48 Veto PMTs

1 SCB 4 Neck PMTs

fast

Test pulser

Readout

deap05

1 V1740 with self-trigger

DTM

4 V1740s

32 V1720s

1 V1720

Online monitoring

Logger

deap00

deapana

Figure 21: Overall DEAP electronics architecture. Shaded boxes are hardwarecomponents; white boxes are software programs. SCBs are signal condition-ing boards to shape PMT signals. V1720s and V1740s are commercial CAENdigitizers. DTM is the trigger module.

50

6.1. Hardware

The DAQ system is housed in 3 computer racks. Each rack has an unin-

terruptible power supply with enough power to safely shut down the system in

case of a power outage underground.

6.1.1. PMTs and Signal Conditioning

In addition to the 255 signal PMTs described in Section 5.1, signals from

the 48 Hamamatsu R1408 8-inch PMTs (45 active) for the outer detector muon

veto, shown in Figure 20, and 4 Hamamatsu R7600-300 PMTs forming a neck

veto (see Section 5.2) are recorded by the DAQ.

The PMTs are powered by a WIENER MPOD crate [62] with ISEG high

voltage modules [63] through 27 custom Signal Conditioning Boards (SCBs).

Each SCB handles up to 12 PMTs, with 22 boards dedicated to inner-detector

PMTs, 4 to the muon veto PMTs, and 1 to the neck veto PMTs. The SCBs

decouple the high voltage, provide high voltage protection, and shape the PMT

signals. The PMT bases are back-terminated with a 4.7 nF capacitor in series

with a 75 Ω resistor, and connected to the SCBs by a 75 Ω impedance cable,

each approximately 20 metres long.

Each SCB has 12 identical channels to shape and amplify the PMT signals.

There are three outputs from each channel: a high-gain channel, a low-gain chan-

nel, and a summing channel. The high-gain channel is designed to achieve high

signal-to-noise for single photoelectrons and shape the pulse to better match the

250 MS/s V1720 digitizer. The low-gain channel is designed to handle pulses

that saturate the high-gain channel and is attenuated by a factor 10 in ampli-

tude. The low-gain pulses are also shaped to be significantly wider, to better

match the 62.5 MS/s V1740 digitizer. The 12 summing channels are added to

create an analog sum (ASUM) for each SCB. The 22 ASUMs from inner detector

SCBs are passed to the DTM with a 24-channel differential connector.

In addition to the 12 safe high voltage (SHV) inputs from the PMTs, each

SCB contains a “test pulse” input. The test pulse is created by the DTM,

and is sent through a discriminator and a fan-out board, to be distributed to

51

Time (ns)6400 6450 6500 6550 6600 6650 6700 6750

AD

C v

alu

e

3850

3860

3870

3880

3890

3900

3910

3920

Figure 22: Example of a measured SPE pulse on a V1720 channel, in ZLE mode.The baseline is set to 3900 ADC, the ZLE threshold for saving data is 3895 ADC,and 20 extra pre-samples and post-samples are saved. Time is in 4 ns binning.

all SCBs simultaneously. Within each SCB, the test pulse is distributed to all

12 channels, with a 0.2 ns channel-to-channel delay. This system allows for easy

extraction of timing offsets between different digitizer channels.

Tagging PMTs for calibration sources are also read out by the DAQ when

deployed. These PMTs are powered by external power supplies, rather than the

SCBs.

6.1.2. Digitizers

The high-gain outputs from the SCBs are connected to 250 MS/s CAEN

V1720 waveform digitizers (8 channels, 12 bits) using MCX cables. V1720s can

store data either in Zero Length Encoding (ZLE) mode or as full waveforms. An

example of an SPE pulse in ZLE mode is shown in Figure 22. The ZLE algorithm

records data only if a given number of samples drop below a threshold ADC

value. The noise level is approximately 1.2 ADC on the V1720 channels, and

a typical SPE pulse is approximately 50 ADC high. The ZLE threshold is set

to be 5 ADC below the baseline of 3900 ADC, providing a balance between

recording real pulses and limiting recorded noise fluctuations. Additionally,

20 extra samples (80 ns) before and after the pulse are recorded. The baseline

is set to 3900 ADC, rather than closer to the maximum 4096 ADC, to allow

overshoot to be recorded.

52

The low-gain outputs from the SCBs are connected to 62.5 MS/s CAEN

V1740 waveform digitizers (64 channels, 12 bits) using MCX cables. The V1740s

do not allow for recording data in ZLE mode, and can only record full waveforms.

Software is used to filter these waveforms so they do not dominate the data rate

(see Section 6.2).

The muon veto PMTs are connected to a V1740 running in “self-trigger”

mode. Instead of the DTM analyzing the veto PMT signals, the digitizer itself

decides whether there is sufficient activity in the water tank to trigger readout.

The 48 PMTs are separated into 6 groups of 8 PMTs, and if any channel in

a group exceeds a height threshold of 15 ADC (approximately 0.75 PE), that

group is deemed to be “active”. If three groups are simultaneously active, then

the self-trigger condition is met. A signal is then sent to the DTM.

The neck veto PMTs are connected to a V1720 running in ZLE mode. During

calibrations with radioactive sources that have tagging PMTs, the signals from

the tagging PMTs are also handled by this V1720.

The digitizers are read out through optical links using proprietary CAEN

A3818 cards [64]. Each card handles 4 optical links, and two V1720s are daisy-

chained on the same link. Each card therefore reads out 8 V1720s, or 4 V1740s.

6.1.3. Trigger

The DTM is responsible for making the trigger decision, providing the mas-

ter clock to synchronize digitizers, triggering digitizers and external calibration

systems, and throttling data collection if the DAQ is overwhelmed. The DTM

hardware is based on a TRIUMF-designed 6U VME motherboard with an AL-

TERA Stratix IV GX field-programmable gate array (FPGA). The motherboard

has three daughterboards connected through FMC standard connectors [65].

The daughterboards are a 24-channel ADC card for digitizing the ASUM chan-

nels from the SCBs [66], a 12-channel NIM I/O card with 8 outputs and 4 in-

puts, and a master clock distribution board. The master clock is distributed at

62.5 MHz to the digitizers, while the ADCs and main FPGA run at 45 MHz.

The NIM outputs are connected to the digitizers, LED light injection system

53

(Section 5.3.1), and a test pulse system.

The trigger system is based on a set of “trigger sources” and logical “trigger

outputs”. Trigger sources, from internal analysis of PMT signals, periodic trig-

gers, or external triggers, are responsible for deciding whether a trigger signal

should be issued. Trigger outputs decide which hardware should be fired. A

logical “trigger output” may fire one, several, or no NIM outputs, and can be

configured to ignore a certain percentage of trigger signals (pre-scaling). Each

trigger source can be connected to many trigger outputs, all configured differ-

ently.

The main trigger algorithm used in DEAP-3600 is the “physics trigger”,

which adds the 22 ASUMs together for a sum of all 255 inner-detector PMTs.

Rolling integrals are computed in two windows, nominally 177 ns and 3100 ns,

aligned to the same start time. The total charge in the prompt window (Eprompt)

and the ratio of energy in the prompt and late windows (Fprompt) are calculated.

The (Eprompt, Fprompt) phase space is split into 6 regions, as shown in Figure 23.

Events in region X are discarded, while the other 5 regions all count as separate

trigger sources. The Eprompt threshold of 1000 ADC gives a 50% detection

efficiency for events with 30 prompt PE. The lower bound of region E is set to

be above the end of the 39Ar beta decay spectrum.

The standard trigger setup for DEAP-3600 data-taking uses the physics

trigger, a periodic trigger, and the muon veto self-trigger. The physics trigger

is set to not read out the digitizers for 99% of events in region C, which is

dominated by 39Ar beta decays. Digitizers are read out for all events in regions

A, B, D and E. Summary information (time, Eprompt and Fprompt) is stored for

all events, regardless of whether the digitizers were read out or not. The periodic

trigger runs at 40 Hz, with test pulses injected at 1 Hz. The remaining 39 Hz

are used to monitor the PMTs, as described in Section 6.3. The veto PMTs are

only read out when the muon veto self-trigger fires. The overall trigger rate is

3200 Hz, with digitizers read out at 500 Hz.

An additional trigger mode is for daily LED light injection calibration, which

only uses periodic triggers. The monitoring trigger is accompanied by a 1 kHz

54

Trigger Eprompt (ADC)

310 410 510 610

Trigger

Fpro

mpt

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1

10

210

310

410

X

A

B

C

D

E

Figure 23: Trigger prompt energy and trigger Fprompt for example data fromthe physics trigger, with no cuts applied. Darker colours indicate more events.The 6 labelled trigger regions are shown in magenta.

periodic trigger that fires the light injection system and digitizers.

6.2. Software and Data Rate Reduction

The readout software is responsible for interfacing with the digitizers and

the DTM, filtering out unnecessary information, collecting information into a

single event, and writing that event to disk. In total, 7 computers are involved

in the readout, all running Scientific Linux 6.6. Four computers handle the inner

detector V1720 data, with additional PCs for the DTM data, the inner detector

V1740 data, the muon veto and neck veto data, and a master.

The V1720 digitizer saturates at approximately 100 PE, and V1740 infor-

mation is only needed for pulses larger than this. Unnecessary V1740 data is

filtered out in two stages. The first stage only relies on V1740 information, and

filters out any waveforms that do not go below 3750 ADC, from a baseline of

3900 ADC. This reduces the V1740 data to be passed to the master computer,

where the second filtering stage removes V1740 waveforms in which the the cor-

responding V1720 channel does not go below 500 ADC. Over 99.9% of V1740

information is removed through this process.

Filtering is also applied to the V1720 waveforms, to further reduce the

amount of data written to disk, and only summary information about the pulse

55

is saved. This summary information is sufficient to give sub-ns timing resolu-

tion of the peak position, as well as the pulse charge, height, and baseline (with

RMS). SPE identification uses probability distribution functions of the ratio of

pulse height to pulse charge, the ratio of pulse height to maximum derivative,

the width, and total charge of the pulse.

After all filtering and compression measures, the data rate is reduced from

7 GB/s to 6 MB/s.

6.3. Operation

The DAQ system is based on the MIDAS package [67], developed by TRI-

UMF and the Paul Scherrer Institute. A web interface has been written to

allow easy remote usage of the DAQ which additionally interfaces with DEAP’s

CouchDB database. The database stores DAQ parameters to be used for given

run types, which are selected by the operator at the start of each run and

automatically forwarded to the appropriate DAQ components.

The DAQ has been designed to run semi-autonomously, with fail-safes in case

of software, hardware or network malfunction. In these scenarios the current run

is stopped and an SMS and email are sent to the DAQ operator. If the operator

does not fix the problem and start a new run within 15 minutes, the PMTs are

ramped down. The entire DAQ system performs a controlled shutdown in case

of power loss or excessive temperature in the DAQ racks due to loss of cooling

water. The data-taking uptime is greater than 95%, excluding unexpected power

outages.

6.4. Database and Data Flow

DEAP uses an Apache CouchDB database [68] to consolidate all external pa-

rameters necessary to configure the DAQ, to calibrate and analyze detector data,

and to evaluate the data quality. Parameter validity ranges are implemented by

run number, and parameter values are version-controlled. Additionally, read-

ings from the DeltaV slow controls system, which monitors physical detector

statuses that are relevant to the analysis, are transferred continuously to a

56

PostgresSQL database. The DEAP analysis framework transparently queries

either the CouchDB or the PostgresSQL database in support of analysis tasks.

ROOT files containing raw detector data are transferred from the DAQ

storage computer to the main analysis cluster. A first-pass analysis translates

DAQ into physics units, applying all necessary calibrations and corrections such

as PMT gains and channel timing offsets. Simultaneously, diagnostic plots for

data quality control are generated, which are available online at the end of the

run. The second analysis pass starts from the calibrated data and calculates

high level information about each event, including event position and number

of detected photoelectrons. The result of each reduction step is written to a

separate set of ROOT files.

The analysis software is maintained in a Git repository on a self-hosted

GitLab server, which provides a web interface and performs automated builds

and tests after every commit (GitLab-CI).

7. Detector Infrastructure

The DEAP-3600 detector is located in the Cube Hall at SNOLAB next to the

MiniCLEAN experiment [69]. A schematic drawing of the detector within the

larger infrastructure is shown in Figure 24. The Cube Hall has approximately

15 × 15 × 15 m3 of useable space and is furnished with an overhead 9-tonne

monorail crane and a 10-tonne gantry crane (Konecranes) on the deck.

The deck is supported by 6 columns and stands over the two experiments to

support the detector, process systems hardware, and electronics. The detector

is housed in a spherical steel shell hanging from a 45-cm-diameter outer neck

rigidly coupled to the central support assembly (CSA). The CSA is connected

through seismic bushings to the deck to allow movement during a seismic event.

The system is designed for survival of the SNOLAB seismic design event, a

Nuttli 4.3 seismic event at 160 m from the site with a peak particle velocity of

800 mm/s [6].

The steel shell, 3.4 m in diameter and fabricated from 304 stainless steel

(All-Weld, Toronto, Canada) with an electropolished inner surface acts as a

57

Figure 24: Model rendering of the DEAP-3600 detector and surrounding infras-tructure inside the Cube Hall at SNOLAB.

water-tight and light-tight vessel, and directs cryogen boil-off to a vent header

in the event of a structural failure in the AV. It was designed as an ASME Section

VIII pressure vessel to withstand the maximum pressure of 30 psig which could

be developed with a fracture in the AV. The entire steel shell was helium leak

tested after construction.

Hanging coaxially inside the steel shell neck is a 30-cm-diameter stainless

steel inner neck 3 m in length, which is coupled to the CSA through a custom

load cell (Sensing Systems) and supports the approximately 13,000 kg inner de-

tector (LAr, AV, LGs, filler blocks, PMTs and detector cabling). The inner neck

was fabricated from a single length of seamless tubing and electropolished. De-

tector cabling and gas purge lines feed along the outside of the inner neck, and

pass through vacuum feedthrough flanges on the CSA. The vapor space between

the inner and outer necks provides the relief path for boil-off argon gas in the

AV failure scenario. The AV bolts to the bottom of the inner neck through an

acrylic-to-steel coupling flange. When cold, the acrylic flange contracts signifi-

cantly more than the metal. To compensate for this, Belleville Spring washers

58

were installed to ensure the bolts maintain sufficient clamping force through

cooling.

The glovebox, shown in Figure 1, is a permanent interface located on the

deck at the top of the neck, used to deploy components to the interior of the

detector in a clean, radon-reduced environment. It is a 76-cm-diameter, 76-cm-

tall cylindrical stainless steel vessel maintained either under vacuum or with a

radon-scrubbed boil-off nitrogen over-pressure purge. The glovebox is equipped

with butyl rubber dry-box gloves for access and inside manipulation. A 45-

cm gatevalve on the top permits loading of large items. Additional 5.5-m and

4.8-m-long cylindrical deployment canisters would be bolted to the gatevalve,

and with the use of an internal hoist were used to extend the clean glovebox

volume for extraction of the resurfacer, and to execute the TPB deposition and

diffuse-laser ball calibration, and to install the neck cooling coil.

The steel shell is surrounded by a cylindrical water shield tank with a di-

ameter and height of 7.8 m made from curved galvanized steel panels. A vinyl

U-shaped liner is located inside the cylinder with an additional vinyl backing

material between the inner liner and tank. A SQ26-10 high sensitivity hy-

drophone (Sensor Technology Ltd.) is installed in the water tank and read out

with the DAQ to provide additional information on seismic events.

The water shield tank has a 75-L/min purification system consisting of a

magnetically coupled pump (IWAKI America MX-251), mixed bed ion exchange

columns (Purolite NRW-37), 0.45-µm filtration (Shelco 3FOS2), degasser (Mem-

brana Liqui-Cel 4x28) and a 254-nm UV sterilizer (Viqua UV Max F4 Plus).

An optional flow path for degassing the water with a nitrogen purge is also

provided. The purification system is run continuously. Two water chillers (Tek-

Temp Instruments, Inc.) maintain the shield tank water at approximately 12C.

Additional chilled water supplies are provided by SNOLAB for cooling of the

DAQ, LN2 cryocoolers, and as-needed auxiliary systems.

59

8. Safety

Large volumes of cryogen are used underground for both the DEAP-3600

and MiniCLEAN experiments. For the safety of underground personnel, over-

pressure protection and oxygen-deficiency analyses were performed.

8.1. Over-pressure Protection

A failure of the DEAP AV would allow the LAr at 87 K to spill onto the

warmer detector components, quickly boiling and releasing a large volume of

inert gas which would lead to an over-pressurization of the steel shell. To miti-

gate this hazard, a burst disk was installed near the top of the steel shell neck

feeding to a 12-inch-diameter vent pipe which services both DEAP-3600 and

MiniCLEAN. It is 335 m long and terminates at a large-capacity mine-air raise.

The 120,000 cfm upward air flow serves to dilute the vented gas and remove it

from the mine. A detailed analytic model was developed to predict the worst-

case vapor generation rate as a function of time. Sample room-temperature

detector components were dipped into LAr and the measured vapor genera-

tion rates were then extrapolated to the full detector. This model predicts a

peak vapor generation rate of 45 kg/s which decreases to less than 5 kg/s after

1 minute.

In addition to the vent pipe, a secondary pressure safety valve provides

pressure relief into the Cube Hall. An 8-inch-diameter pipe connects the CSA

to a rupture disk set at 15 psig followed by a check valve set at 5 psig. The

main vent pipe can provide a steady-state flow of 13 kg/s of argon gas with

a pressure drop of 15 psig. An analytic model of the transient gas flow using

the worst-case vapor generation rate predicts that the secondary pressure safety

valve would open for approximately 80 s and release 220 kg of argon gas into

the Cube Hall.

8.2. Oxygen Deficiency

Analysis of the oxygen deficiency hazard (ODH) followed the methodol-

ogy developed at Fermi National Accelerator Laboratory [70] (Fermilab), which

maintains tables of failure rates for many components used in cryogenic systems.

60

Several techniques have been implemented to mitigate the oxygen deficiency

hazard. The large cryogen storage vessels are certified ASME section VIII pres-

sure vessels, resulting in a low probability of failure. Additionally, the outlet of

the nitrogen storage vessel has a flow-restriction of 100 g/s. Seismic isolation

pads are used for the argon dewar, nitrogen dewar, and steel shell, and mechan-

ical barriers are in place to prevent contact with dewars by heavy equipment.

Due to the large volume of the Cube Hall, mixing of the air is sufficient to

dilute vented gas in all scenarios. In addition to the main SNOLAB ventilation

system, a secondary air mixing system was installed. This consisted of three

5000 cfm and two of 300 cfm fans on the Cube Hall floor, two 300 cfm fans on

the deck, and two 1000 cfm fans at the top of the Cube Hall near the nitrogen

dewar. These run at all times and the supplied currents are monitored and

backed up by an uninterruptible power supply. A series of oxygen monitors

(PureAire TX-1100-DRA) are installed around the Cube Hall and in adjacent

halls.

Spill tests, in which 100 kg to 150 kg of argon were flash boiled on the Cube

Hall floor in less than 30 seconds, generated oxygen levels below 135 mmHg,

not safe for personnel, at the Cube Hall floor monitors. The oxygen levels were

extrapolated to the model failure scenario. From this, a set of four 2500 cfm

on-demand vertical mixing fans (Pearson 12 inch Velocity), triggered when the

oxygen sensors read low, was installed to move air from the Cube Hall floor to

the ceiling to promote mixing and dilution.

With all mitigation steps in place, the probability of a serious accident using

the Fermilab methodology is below 10−7 hr−1 which places the Cube Hall in a

category that does not require staff to routinely have access to a re-breather,

self-contained breathing apparatus, or other emergency air systems.

9. Summary

The DEAP-3600 detector searches for dark matter particle interactions us-

ing single-phase liquid argon technology. The projected WIMP-nucleon cross-

section sensitivity for a 3-tonne-year fiducial exposure is 10−46 cm2 at 100 GeV/c2

61

WIMP mass.

The use of a simple, single-phase liquid argon target contained in an acrylic

cryostat is novel. Significant design, research, and development have been un-

dertaken to minimize detector backgrounds including quality control during the

ultra-pure acrylic vessel manufacturing and resurfacing, the selection of low ra-

dioactivity materials, and limiting the exposure of detector components to radon

during assembly and construction.

The PMT system has been operational since the end of 2014. The cryogenic

handling and purification system was commissioned in winter 2014. The water

shield tank components, application of the wavelength shifter, calibration hard-

ware and muon veto PMT system were completed in summer 2015. Installation

of the final argon delivery system occurred in the fall of 2015, with cool-down

of the acrylic vessel in the spring of 2016. The LAr fill began in the summer

of 2016. After the neck seal failure incident on 17 August 2016, the AV was

emptied of the LAr and refilled to a reduced level. Stable operations continue

with a LAr mass of 3260 kg.

Acknowledgements

This work is supported by the Natural Sciences and Engineering Research

Council of Canada (NSERC), the Canadian Foundation for Innovation (CFI),

the Ontario Ministry of Research and Innovation, and Alberta Advanced Edu-

cation and Technology (ASRIP). We acknowledge support from the European

Research Council Project ERC StG 279980, the UK Science & Technology Facil-

ities Council (STFC) grant ST/K002570/1, the Leverhulme Trust grant number

ECF-20130496, the Rutherford Appleton Laboratory Particle Physics Division,

and STFC and SEPNet PhD studentship support. We acknowledge support

from DGAPA-UNAM through grant PAPIIT No. IA100316. We thank Com-

pute Canada, Calcul Quebec, McGill University’s centre for High Performance

Computing and the Center for Advanced Computing (CAC) for computational

support and data storage.

62

We additionally thank SNOLAB and its staff for support through under-

ground space, logistical, and technical services. SNOLAB operations are sup-

ported by the Canada Foundation for Innovation and the Province of Ontario

Ministry of Research and Innovation, with underground access provided by Vale

at the Creighton mine site. We thank Vale for their transportation of the acrylic

vessel from surface to SNOLAB. On-site construction could not have been com-

pleted without the underground contractors, undergraduate research associates,

and summer and co-op students who have made enormous contributions, and the

management of Tony Flower. We thank the following people for their valuable

inputs: David Bearse (Queen’s), Neil Tennyson (Alfa Aesar), Dan Runciman

(Johnston Industrial Plastics), Carlos Guerra (Spartech), Kalayil T. Varghese

(RPT Asia) and Micha l Tarka (Stony Brook).

References

[1] G. Jungman, M. Kamionkowski, K. Griest, Supersymmetric dark matter,

Physics Reports 267 (5-6) (1996) 195–373.

[2] G. Bertone, D. Hooper, J. Silk, Particle dark matter: Evidence, candidates

and constraints, Physics Reports 405 (5) (2005) 279–390.

[3] E. Aprile et al. (Xenon Collaboration), First dark matter search results

from the XENON1T experiment, arXiv preprint arXiv:1705.06655.

[4] P. Agnes et al. (DarkSide Collaboration), Results from the first use of low

radioactivity argon in a dark matter search, Physical Review D 93 (8)

(2016) 081101.

[5] Z. Ahmed et al. (CDMS II Collaboration), Dark matter search results from

the CDMS II experiment, Science 327 (5973) (2010) 1619–1621.

[6] F. Duncan, A. Noble, D. Sinclair, The construction and anticipated science

of SNOLAB, Annual Review of Nuclear and Particle Science 60 (1) (2010)

163.

63

[7] M. Boulay et al. (DEAP Collaboration), Measurement of the scintillation

time spectra and pulse-shape discrimination of low-energy beta and nuclear

recoils in liquid argon with DEAP-1, Astroparticle Physics 85 (2016) 1–23.

[8] M. Kuzniak et al. (DEAP Collaboration), DEAP-3600 dark mat-

ter search, Nuclear and Particle Physics Proceedings (2016)

273doi:http://dx.doi.org/10.1016/j.nuclphysbps.2015.09.048.

URL http://www.sciencedirect.com/science/article/pii/

S2405601415005374

[9] B. Broerman et al., Application of the TPB wavelength shifter to the

DEAP-3600 spherical acrylic vessel inner surface, Journal of Instrumen-

tation 12 (04) (2017) P04017.

URL http://stacks.iop.org/1748-0221/12/i=04/a=P04017

[10] T. Doke et al., Absolute scintillation yields in liquid argon and xenon for

various particles, Japanese journal of applied physics 41 (3R) (2002) 1538.

[11] D. Gastler et al., Measurement of scintillation efficiency for nuclear recoils

in liquid argon, Physics Review C 85 (2012) 065811.

[12] O. Cheshnovsky, B. Raz, J. Jortner, Emission spectra of deep impurity

states in solid and liquid rare gas alloys, The Journal of Chemical Physics

57 (11) (1972) 4628–4632.

[13] V. Gehman et al., Fluorescence efficiency and visible re-emission spec-

trum of tetraphenyl butadiene films at extreme ultraviolet wavelengths,

Nuclear Instruments and Methods in Physics Research Section A: Acceler-

ators, Spectrometers, Detectors and Associated Equipment 654 (1) (2011)

116–121.

[14] Hamamatsu R5912 photomultiplier tube datasheet.

URL http://www.hamamatsu.com/resources/pdf/etd/LARGE_AREA_

PMT_TPMH1286E.pdf

64

[15] J. Oetel, R. Bourque, C. Hirschi, F. Lopez, The design criteria and engi-

neering of a vacuum cryostat and inner acrylic vessel for a noble liquid dark

matter detector, LANL Internal Document. LA-UR-07-8170 (2007).

[16] P. Benetti et al., Measurement of the specific activity of 39Ar in natural

argon, Nuclear Instruments and Methods in Physics Research Section A:

Accelerators, Spectrometers, Detectors and Associated Equipment 574 (1)

(2007) 83–88.

[17] M. G. Boulay, A. Hime, Technique for direct detection of weakly interacting

massive particles using scintillation time discrimination in liquid argon,

Astropart. Phys. 25 (2006) 179.

[18] T. Caldwell, Simulation of Noble Liquid Detectors Using RAT, presented

at AARM Meeting at Fermilab, March 19-22 (2014).

[19] S. Agostinelli et al., Geant4: A simulation toolkit, Nuclear Instruments

and Methods in Physics Research Section A: Accelerators, Spectrometers,

Detectors and Associated Equipment 506 (3) (2003) 250 – 303. doi:https:

//doi.org/10.1016/S0168-9002(03)01368-8.

[20] R. Brun, F. Rademakers, Root: an object oriented data analysis framework,

Nuclear Instruments and Methods in Physics Research Section A: Accelera-

tors, Spectrometers, Detectors and Associated Equipment 389 (1-2) (1997)

81–86.

[21] M. Bodmer et al., Measurement of optical attenuation in acrylic light

guides for a dark matter detector, Journal of Instrumentation 9 (02) (2014)

P02002.

[22] M. Janecek, W. W. Moses, Measuring light reflectance of BGO crystal

surfaces, Nuclear Science, IEEE Transactions on 55 (2008) 2443–2449.

[23] T. Pollmann, M. Boulay, M. Kuzniak, Scintillation of thin tetraphenyl bu-

tadiene films under alpha particle excitation, Nucl. Instrum. Methods A

635 (1) (2011) 127.

65

[24] L. Veloce, An investigation of backgrounds in the DEAP-3600 dark matter

direct detection experiment, Master’s thesis, Queen’s University (2013).

[25] L. Veloce, M. Kuzniak, P. Di Stefano, A. Noble, M. Boulay, P. Nadeau,

T. Pollmann, M. Clark, M. Piquemal, K. Schreiner, Temperature depen-

dence of alpha-induced scintillation in the 1, 1, 4, 4-tetraphenyl-1, 3-

butadiene wavelength shifter, Journal of Instrumentation 11 (06) (2016)

P06003.

[26] P.-A. Amaudruz et al. (DEAP Collaboration), Radon backgrounds in the

DEAP-1 liquid-argon-based dark matter detector, Astropart. Phys. 62

(2015) 178.

[27] M. Kuzniak, M. G. Boulay, T. Pollmann, Surface roughness interpretation

of 730 kg days CRESST-II results, Astropart. Phys. 36 (1) (2012) 77.

[28] M. H. Mendenhall, R. A. Weller, Algorithms for the rapid computation

of classical cross sections for screened coulomb collisions, Nucl. Instrum.

Methods B 58 (1991) 11.

[29] M. H. Mendenhall, R. A. Weller, An algorithm for computing screened

coulomb scattering in Geant4, Nucl. Instrum. Methods B 227 (2004) 420.

[30] J. F. Ziegler, M. D. Ziegler, J. P. Biersack, SRIM–the stopping and range

of ions in matter (2010), Nuclear Instruments and Methods in Physics

Research Section B: Beam Interactions with Materials and Atoms 268 (11)

(2010) 1818–1823.

[31] D.-M. Mei, Z.-B. Yin, L. Stonehill, A. Hime, A model of nuclear recoil

scintillation efficiency in noble liquids, Astropart. Phys. 30 (1) (2008) 12 –

17.

[32] M. Szydagis et al. (NEST Collaboration), NEST: a comprehensive model

for scintillation yield in liquid xenon, Journal of Instrumentation 6 (10)

(2011) P10002.

66

[33] A. Lindote, H. Arajo, V. Kudryavtsev, M. Robinson, Simulation of neu-

trons produced by high-energy muons underground, Astropart. Phys. 31 (5)

(2009) 366 – 375.

[34] M. Horn, Simulations of the muon-induced neutron background of the

EDELWEISS-II experiment for dark matter search, Ph.D. thesis, Karl-

sruher Institut fur Technologie (2007).

URL http://nbn-resolving.org/urn:nbn:de:swb:90-74021

[35] W. Wilson, R. Perry, W. Charlton, T. Parish, Sources: A code for cal-

culating (alpha, n), spontaneous fission, and delayed neutron sources and

spectra, Progress in Nuclear Energy 51 (45) (2009) 608 – 613. doi:http:

//dx.doi.org/10.1016/j.pnucene.2008.11.007.

[36] S. Westerdale, P. Meyers, Radiogenic neutron yield calculations for low-

background experiments, arXiv preprint arXiv:1702.02465.

[37] J. Xu et al., A study of the residual 39Ar content in argon from underground

sources, arXiv preprint arXiv:1204.6011.

[38] Hitachi, Nanocrystalline soft magnetic material, FINEMET.

URL http://www.hilltech.com/pdf/hl-fm10-cFinemetIntro.pdf

[39] I. Lawson, Ultra-low background measurement capabilities at SNOLAB,

in: Journal of Physics: Conference Series, Vol. 718, IOP Publishing, 2016,

p. 042034.

[40] Styrofoam TM brand SM extruded polystyrene foam insulation.

URL http://www.dowbuildingsolutions.com

[41] M. Liu, H. Lee, A. McDonald, 222Rn emanation into vacuum, Nuclear In-

struments and Methods in Physics Research Section A: Accelerators, Spec-

trometers, Detectors and Associated Equipment 329 (1-2) (1993) 291–298.

[42] H. Simgen, Ultrapure gases for GERDA, Progress in Particle and Nuclear

Physics 57 (1) (2006) 266–268.

67

[43] C. M. Nantais, Radiopurity measurement of acrylic for the DEAP-3600

dark matter experiment, Master’s thesis, Queen’s University (2014).

[44] T. Araki et al. (KamLAND Collaboration), Measurement of neutrino os-

cillation with KamLAND: Evidence of spectral distortion, Physical Review

Letters 94 (8) (2005) 081801.

[45] C. Jillings, Control of contamination of radon-daughters in the DEAP-

3600 acrylic vessel, in: Low Radioactivity Techniques 2013 (LRT 2013):

Proceedings of the IV International Workshop in Low Radioactivity Tech-

niques, Vol. 1549, AIP Publishing, 2013, pp. 86–89.

[46] Durridge RAD7 radon detector.

URL http://www.durridge.co.uk/documentation/RAD7%20Manual.pdf

[47] M. Wojcik, Measurement of radon diffusion and solubility constants in

membranes, Nuclear Instruments and Methods in Physics Research Section

B: Beam Interactions with Materials and Atoms 61 (1) (1991) 8–11.

[48] Stirling SPC-1 cryocooler.

URL https://www.stirlingcryogenics.com/files/__documents/1/

SPC-1Datasheet.pdf

[49] DeltaV distributed control system.

URL http://www.emerson.com

[50] J. Stachiw, M. T. Society, Handbook of acrylics for submersibles, hyper-

baric chambers, and aquaria, Best Pub. Co., 2003.

[51] M. Wojcik, G. Zuzel, Low 222Rn nitrogen gas generator for ultra-low back-

ground counting systems, Nuclear Instruments and Methods in Physics

Research Section A: Accelerators, Spectrometers, Detectors and Associ-

ated Equipment 539 (1) (2005) 427–432.

[52] T. Pollmann, Alpha backgrounds in the DEAP dark matter search experi-

ment, Ph.D. thesis, Queen’s University (2012).

68

[53] B. Broerman, On the development of the wavelength shifter deposition

system for the DEAP-3600 dark matter search experiment, Master’s thesis,

Queen’s University (2015).

[54] P.-A. Amaudruz et al. (DEAP Collaboration), In-situ characterization

methods for the Hamamatsu R5912 photomultiplier tubes used in the

DEAP-3600 experiment, submitted to JINST Preprint arxiv:1705.10183.

[55] Radia, Multiplatform software for 3d magnetostatics computation.

URL http://www.esrf.eu/Accelerators/Groups/InsertionDevices/

Software/Radia

[56] B. A. Moffet et al. (SNO Collaboration), Optical calibration for the Sud-

bury Neutrino Observatory, Nucl. Instrum. Methods A 554 (2005) 255–265.

[57] J. S. Kapustinsky et al., A fast timing light pulser for scintillation detectors,

Nuclear Instruments and Methods in Physics Research A 241 (1985) 612.

[58] Hilger-Crystals.

URL http://www.hilger-crystals.co.uk

[59] Hamamatsu Photonics.

URL http://www.hamamatsu.com

[60] CAEN, Technical Information Manual: MOD. V1720 8 Channel 12 Bit 250

MS/S Digitizer Manual Rev. 26.

URL http://www.caen.it/servlet/checkCaenManualFile?Id=10276

[61] CAEN, Technical Information Manual: MOD. V1740 64 Channel 12 Bit 65

MS/S Digitizer Manual Rev. 17.

URL http://www.caen.it/servlet/checkCaenManualFile?Id=9867

[62] WIENER, Plein & Baus, Ltd., MPOD LV/HV System Startup Informa-

tion.

URL http://www.iseg-hv.com/download.php/991/file_url_en/MPOD%

20information.pdf

69

[63] ISEG, EDS Distributor HV Modules with Common-GND Operators

Manual.

URL http://www.iseg-hv.com/download.php/1062/file_url_en/

iseg_EDS_F0_200.pdf

[64] CAEN, Technical Information Manual: A3818 PCI Express Optical Link.

URL http://www.caen.it/servlet/checkCaenManualFile?Id=9775

[65] VITA, ANSI/VITA 57.1 FPGA Mezzanine Card (FMC) Standard.

URL http://www.vita.com

[66] Texas Instruments, ADC12EU050 Ultra-Low Power, Octal, 12-bit, 45

MSPS Sigma-Delta Analog-to-Digital Converter.

URL http://www.ti.com/lit/gpn/adc12eu050

[67] S. Ritt, P. Amaudruz, K. Olchanski, MIDAS documentation.

URL http://ladd00.triumf.ca/~daqweb/doc/midas/html/

[68] Database using JSON for documents.

URL http://couchdb.apache.org/

[69] K. Rielage et al. (MiniCLEAN Collaboration), Update on the MiniCLEAN

dark matter experiment, Physics Procedia 61 (2015) 144–152.

[70] R. Schmitt, FESHM 5064: Fermilab ES&H manual, oxygen deficiency haz-

ards, Tech. rep., Fermilab (2012).

70